Professor Peter Gibbens

Professor Peter Gibbens

Professor

School of Engineering

Career Summary

Biography

joined UON as Professor of Aerospace Systems Engineering at The University of Newcastle in August 2019. My field speciality is in flight mechanics, covering aircraft stability and control and handling qualities, flight control systems, guidance and navigation systems, avionics and flight simulation.

My research interests are in advanced control methods and visual navigation for next generation Electronic Flight Information Systems (EFIS). I have spent two decades developing methods in Model Predictive Control that provide efficient and accurate prediction for precise pre-emptive control, especially in the application of these methods to gust alleviation in flight. My research into visual navigation involves navigation relative to the visual environment by detecting and tracking natural and man-made features in the environment using optical sensors. The objective is to enhance current generation EFIS systems that have Synthetic Vision with a loop closure that automatically correlates visual features on the ground with their registrations in a database (or on a map) to fix the aircraft’s position, thus mimicking pilot behaviour in VFR (Visual Flight Rules) flight. These approaches combine image processing and feature characterisation with data fusion techniques for navigation. In particular, a blend of Simultaneous Localisation and Mapping (SLAM) and Terrain Aided Navigation (TAN) is used to provide an optimal relative navigation solution. These techniques have a natural extension to use in UAV’s and space robotics applications operating in unstructured and GPS denied environments.

I have strong interest in engineering education and effective training via experiential learning. I have won several local and national teaching awards for the implementation of Variable Stability Flight Simulation in coursework to improve learning of aircraft stability, response and handling qualities, flight control system design and implementation, and flight testing for aerodynamic identification.

Prior to joining the University of Newcastle, I spent 23 years at the University of Sydney in Aerospace Engineering, and earlier worked with the Defence Science and Technology Organisation in their Air Operations and Flight Mechanics Divisions. This work involved flight simulation and aerodynamic modelling, flight stability and control and mission analysis, and flight testing for aerodynamic identification, principally on aircraft in Australia’s defence inventory. I was involved in the flight test programme of F111-C to characterise the aerodynamics of the aircraft from dynamic manoeuvres and to validate simulation programmes used in mission analysis, accident reconstruction and the F111-C mission flight simulator upgrade.

My current role at the University of Newcastle is to design and establish the new degree in Aerospace Systems Engineering. The degree was initiated in response to requests from

Aerospace industry partners at Williamtown Air Base supporting the new Joint Strike Fighter. The goal is to establish an aerospace engineering degree focussing on the engineering of aerospace systems, taking a systems-of-systems approach in which the aircraft is but one subsystem. The focus is to provide an education covering the breadth of systems and subsystems involve in an “aircraft” system. Although the degree will cover elements involved in classical aerospace engineering involving the design and analysis of aerospace vehicles, the broader goal is to give students an appreciation and a working knowledge in all of the subsystems that make an aerospace system work. This involves airframe, propulsion systems, power and communications systems, sensors, control systems and avionics, as well as ground and space-based support infrastructure. The inherent theme is systems integration, and the degree content will implicitly embed in students the generic skills involved in Systems Engineering. The degree will involve experiential learning components wherever possible and I am working to establish an aerospace systems teaching laboratory that will place UON at the forefront of aerospace systems teaching, learning and research nationally.


Qualifications

  • Doctor of Philosophy, University of Newcastle
  • Bachelor of Engineering in Aeronautical Engineering, University of Sydney

Keywords

  • aerodynamic parameter estimation
  • aircraft systems and avionics
  • automatic flight control
  • flight dynamics
  • simultaneous localisation and mapping
  • variable stability flight simulation

Languages

  • English (Mother)

Fields of Research

Code Description Percentage
090104 Aircraft Performance and Flight Control Systems 40
090106 Flight Dynamics 20
090904 Navigation and Position Fixing 40

Professional Experience

UON Appointment

Title Organisation / Department

Academic appointment

Dates Title Organisation / Department
1/1/1996 - 23/8/2019 Assiciate Professor in Aerospace Engineering The University of Sydney
Aerospace, Mechanical and Mechatronic Engineering
Australia

Professional appointment

Dates Title Organisation / Department
1/10/1985 - 30/4/1988 Experimental Officer Defence Science & Technology Organisation
Flight Mechanics Division
Australia
1/7/1992 - 31/12/1995 Research Scientist Defence Science & Technology Organisation
Air Operations Division
Australia
1/5/1988 - 30/6/1992 Research Scientist (Cadet) Defence Science & Technology Organisation
Flight Mechanics Division
Australia

Awards

Award

Year Award
2018 Dean's Faculty Award for Outstanding Teaching Innovation in 2017
University of Sydney
2010 Citation for Outstanding Contribution to Student Learning
Australian Learning and Teaching Council
2010 Vice-Chancellor's Award for Outstanding Teaching
University of Sydney
2009 Vice-Chancellor's Award for Outstanding Teaching 2009
University of Sydney
2009 Dean's Faculty Teaching Award 2009
University of Sydney
1995 AMRL Achievement Award for Outstanding Contribution to Defence
Defence Science and Technology Organisation

Teaching

Code Course Role Duration
AERO4560 Flight Mechanics 2
The University of Sydney
Aircraft automatic flight control systems. Transfer functions and response to flight controls in the frequency domain. Aircraft response to stochastic inputs (gusts). Flight control system design using classical flight control methods.
Lecturer 1/1/1996 - 23/8/2019
AERO3560 Flight Mechanics 1
The University of Sydney
Aircraft flight dynamics, stability and controllability. Aircraft handling qualities. Flight simulation.
Lecturer 1/1/1996 - 23/8/2019
AERO5500 Advanced Flight Mechanics, Test and Evaluation
The University of Sydney
Flight test techniques. Aerodynamic parameter estimation. Modern flight control techniques. Aircraft state estimation. Model validation and verification.
Lecturer 1/1/1996 - 23/8/2019
Edit

Publications

For publications that are currently unpublished or in-press, details are shown in italics.


Chapter (1 outputs)

Year Citation Altmetrics Link
2008 Tsai AC, Gibbens PW, Hugh Stone R, 'Visual position estimation for automatic landing of a tail-sitter vertical takeoff and landing unmanned air vehicle', Mechatronics and Machine Vision in Practice 181-191 (2008)

People gain a physical sense of the environment surrounding them via visual information from the eyes; but from these observations alone we are not capable of determining the exac... [more]

People gain a physical sense of the environment surrounding them via visual information from the eyes; but from these observations alone we are not capable of determining the exact dimensions of the environment. Field robots use sensors such as Global Position System (GPS) or Inertial Measurement Units (IMU) to make accurate estimations of the position and attitude, but these instruments cannot provide accurate relative measurements with respect to a specific site without prior surveying. Computer vision techniques, i. e. using cameras as sensors; offer vision information that gives a physical sense of a robotic platform pose with respect to some targeted site, and are capable of making accurate estimates of relative positions and attitudes. © 2008 Springer-Verlag.

DOI 10.1007/978-3-540-74027-8_14
Citations Scopus - 1

Journal article (17 outputs)

Year Citation Altmetrics Link
2018 Chamitoff GE, Saenz-Otero A, Katz JG, Ulrich S, Morrell BJ, Gibbens PW, 'Real-time maneuver optimization of space-based robots in a dynamic environment: Theory and on-orbit experiments', Acta Astronautica, 142 170-183 (2018)

© 2017 IAA This paper presents the development of a real-time path-planning optimization approach to controlling the motion of space-based robots. The algorithm is capable of plan... [more]

© 2017 IAA This paper presents the development of a real-time path-planning optimization approach to controlling the motion of space-based robots. The algorithm is capable of planning three dimensional trajectories for a robot to navigate within complex surroundings that include numerous static and dynamic obstacles, path constraints and performance limitations. The methodology employs a unique transformation that enables rapid generation of feasible solutions for complex geometries, making it suitable for application to real-time operations and dynamic environments. This strategy was implemented on the Synchronized Position Hold Engage Reorient Experimental Satellite (SPHERES) test-bed on the International Space Station (ISS), and experimental testing was conducted onboard the ISS during Expedition 17 by the first author. Lessons learned from the on-orbit tests were used to further refine the algorithm for future implementations.

DOI 10.1016/j.actaastro.2017.10.001
Citations Scopus - 2
2018 Volkova A, Gibbens PW, 'More Robust Features for Adaptive Visual Navigation of UAVs in Mixed Environments: A Novel Localisation Framework', Journal of Intelligent and Robotic Systems: Theory and Applications, 90 171-187 (2018)

© 2017, Springer Science+Business Media B.V. In this paper, we present an autonomous visual navigation system that determines the location of the unmanned aerial vehicle (UAV) in ... [more]

© 2017, Springer Science+Business Media B.V. In this paper, we present an autonomous visual navigation system that determines the location of the unmanned aerial vehicle (UAV) in GPS-denied environment by detecting semantic features (roads centrelines, intersections, outlines of forest and river) in aerial imagery and matching them to a pre-built dataset. This work is centred around testing the capability of a road centreline modelling and matching algorithm to localise accurately. Alongside, dynamic feature modelling and minimalistic description to optimise data association are proposed. We test three novel datasets with satellite imagery covering the same rural area with significant seasonal and lighting variation.

DOI 10.1007/s10846-017-0650-2
Citations Scopus - 1
2018 Volkova A, Gibbens PW, 'Aerial wide-area motion imagery registration using automated multiscale feature selection', IEEE Geoscience and Remote Sensing Letters, 15 1620-1624 (2018)

© 2004-2012 IEEE. Automatic registration of aerial wide-area motion imagery is required to correct the camera parameters in order to eliminate the geocoding errors arising from fr... [more]

© 2004-2012 IEEE. Automatic registration of aerial wide-area motion imagery is required to correct the camera parameters in order to eliminate the geocoding errors arising from frequent reinstallation of the camera array on the carrier aircraft. Approaches developed to date solely rely on the information present in the imagery not using a priori knowledge about the environment and the features present in it in the sequence analysis. To this end, we propose a novel method based on dynamic feature extraction and automatic multiscale feature matching to produce per-frame camera pose corrections. The features are extracted from the imagery using one of the three dedicated classifiers and are then robustly matched to the features projected from the datum using a coarse-to-fine multiscale approach. Finally, the bias between the estimated and the actual camera pose is estimated using ordinary least squares optimization performed on the distances between the approved match candidate pairs. The application of the proposed method to 50 frames of very high-resolution aerial imagery captured over mixed terrain at an altitude of 5.3 km demonstrates significant reduction in position error of the features (from 47.76 to 12.31 m) and proves the attractiveness of the approach as an alternative to manual labeling methods using ground control points.

DOI 10.1109/LGRS.2018.2841845
Citations Scopus - 2
2016 Lamburn D, Gibbens P, Dumble S, 'Explicit efficient constrained model predictive control', International Journal of Automation and Control, 10 329-355 (2016)

Copyright © 2016 Inderscience Enterprises Ltd. A more efficient model predictive control algorithm is algebraic model predictive control. The efficiency is improved by forming the... [more]

Copyright © 2016 Inderscience Enterprises Ltd. A more efficient model predictive control algorithm is algebraic model predictive control. The efficiency is improved by forming the prediction horizon with a non-uniform distribution of points. The most efficient case reduces the problem down to a single prediction point, defining the horizon length as well as removing any redundant control calculations. This paper shows that for this single prediction point case, the constrained control can be easily evaluated explicitly, thus making the control calculation deterministic and more efficient. We also show that because an explicit control can be determined for the system, the closed-loop constrained stability bounds can be evaluated a priori. Using the key constraint of the peak location of the response also allows the system to meet the constraints while still being explicit and deterministic. The proposed explicit algorithm is applied to two numerical simulation examples with the performance, computational efficiency and stability analysed.

DOI 10.1504/IJAAC.2016.079538
Citations Scopus - 2
2015 Dumble SJ, Gibbens PW, 'Airborne Vision-Aided Navigation Using Road Intersection Features', Journal of Intelligent and Robotic Systems: Theory and Applications, 78 185-204 (2015)

© 2014, Springer Science+Business Media Dordrecht. Modern airborne navigation systems for manned and unmanned platforms usually rely on GPS measurements to constrain the inertial ... [more]

© 2014, Springer Science+Business Media Dordrecht. Modern airborne navigation systems for manned and unmanned platforms usually rely on GPS measurements to constrain the inertial position estimate of the platform. This reliance on GPS can quickly cause the navigational estimates of system to become unreliable when the system is operating in GPS-limited or denied areas. This paper presents a vision-aided inertial navigation system that uses ground features (in this case road intersections) matched to a database to provide position measurements. An image processing algorithm is used to extract the shapes of the road intersections from visual imagery, this shape is then matched to a reference database to provide image to map road intersection correspondences. The correspondence information is fused with the inertial solution in an Extended Kalman Filter to constrain the complete attitude and position inertial navigation solution. The system is developed to operate at non-zero attitude angles removing the level flight limitations of past approaches. Flight test results of the system demonstrate that the system can successfully produce accurate navigation estimates that are comparable to the use of GPS without the same limitations of GPS when operating in a GPS-limited or denied area.

DOI 10.1007/s10846-014-0056-3
Citations Scopus - 6
2015 Dumble SJ, Gibbens PW, 'Efficient Terrain-Aided Visual Horizon Based Attitude Estimation and Localization', Journal of Intelligent and Robotic Systems: Theory and Applications, 78 205-221 (2015)

© 2014, Springer Science+Business Media Dordrecht. Inertial Navigation Systems typically rely on aiding-sensors such as GPS (Global Positioning System) to estimate the location of... [more]

© 2014, Springer Science+Business Media Dordrecht. Inertial Navigation Systems typically rely on aiding-sensors such as GPS (Global Positioning System) to estimate the location of the system. The navigational performance of the Inertial Navigation System can be severely degraded when the GPS measurements are inaccurate or unavailable. Terrain-Aided Navigation is another method of localizing the platform by correlating the measured terrain information with a Digital Terrain Model. This paper presents an efficient Terrain-Aided Navigation method of generating position measurements from the visual appearance of the horizon profile (and hence terrain) surrounding the platform. An optimization process is used to align the measured horizon profile to an off-line pre-generated terrain-aided reference profile which allows for efficient position and attitude estimation. Numerical simulations are presented to evaluate the effectiveness of the proposed method. These results show that precise real-time attitude and position estimation is achievable using visual horizon profile information.

DOI 10.1007/s10846-014-0043-8
Citations Scopus - 14
2014 Medagoda EDB, Gibbens PW, 'Multiple horizon model predictive flight control', Journal of Guidance, Control, and Dynamics, 37 946-951 (2014)

A general multiple horizon predictive control scheme extending the Algebraic Model Predictive Control (AMPC) algorithm in its application to flight control was proposed. It was sh... [more]

A general multiple horizon predictive control scheme extending the Algebraic Model Predictive Control (AMPC) algorithm in its application to flight control was proposed. It was shown that by allocating prediction horizons tuned specifically to the time scales over which the outputs react, the overall response can be greatly improved. The method's ability to decouple output responses of a highly coupled longitudinal aircraft system was also demonstrated, highlighting the additional tuning capabilities that are lost when a single, global prediction horizon is used. The results demonstrate the AMPC algorithms ability in controlling multiple outputs of the linear, longitudinal aircraft model. The performance of the algorithm will be assessed by analyzing the aircraft's response in tracking a commanded climb rate while regulating forward airspeed.

DOI 10.2514/1.62889
Citations Scopus - 7
2014 Lamburn DJ, Gibbens PW, Dumble SJ, 'Efficient constrained model predictive control', European Journal of Control, 20 301-311 (2014)

© 2014 European Control Association. Published by Elsevier Ltd. All rights reserved. Algebraic model predictive control is a more efficient predictive control algorithm in which t... [more]

© 2014 European Control Association. Published by Elsevier Ltd. All rights reserved. Algebraic model predictive control is a more efficient predictive control algorithm in which the optimal control problem can be reduced down to a single prediction point, thus removing any additional and redundant calculations. The drawback of the reduction is lost system information and hence degraded performance while the system is constrained. This paper introduces a method of improving the constraint handling performance of the system by introducing additional points at which the constraints are checked. The computational efficiency is retained by recognizing that the key constraint is the peak location of the response, with a method shown for calculating its location. The proposed method is demonstrated on a numerical simulation of a second order system and linear aircraft model with performance and computational efficiency compared.

DOI 10.1016/j.ejcon.2014.08.001
Citations Scopus - 3
2012 Dumble SJ, Gibbens PW, 'Horizon profile detection for attitude determination', Journal of Intelligent and Robotic Systems: Theory and Applications, 68 339-357 (2012)

The horizon appearance is a strong visual indication of the attitude of an aircraft, so a vision based system should be able to detect the horizon and use its appearance to extrac... [more]

The horizon appearance is a strong visual indication of the attitude of an aircraft, so a vision based system should be able to detect the horizon and use its appearance to extract attitude measurements. Past methods have made the assumption that the horizon is straight, this neglects possible navigational and attitude information. This paper outlines a horizon detection method which allows for the actual horizon profile shape to be extracted. This horizon profile is then used for visual attitude determination. Test results for a captured flight video are presented and the proposed method is compared and evaluated against other methods. © Springer Science+Business Media B.V. 2012.

DOI 10.1007/s10846-012-9684-7
Citations Scopus - 14
2011 Gibbens PW, Medagoda EDB, 'Efficient model predictive control algorithm for aircraft', Journal of Guidance, Control, and Dynamics, 34 1909-1915 (2011)

The use of state-space models has gained increased attention for their ability to easily and accurately control multivariable processes within an MPC framework. An analysis of the... [more]

The use of state-space models has gained increased attention for their ability to easily and accurately control multivariable processes within an MPC framework. An analysis of the controller on a linear longitudinal aircraft model will be performed using a variety of controller configurations, assessing the effectiveness and controllability of the system using the proposed. By using this formulation of the matrix exponential, no approximations are needed, as a direct time-domain solution of the state transition matrix can be obtained. The omission of the higher-order terms in this expansion is an acceptable approximation if the discretization period is sufficiently small. However, if larger discretization periods are used, the higherorder terms become more significant and affect the accuracy of the result. The advantage of using unevenly distributed time intervals is that there is no longer a restriction on where prediction points can be placed or when predicted outputs are to be evaluated.

DOI 10.2514/1.52162
Citations Scopus - 23
2010 Medagoda EDB, Gibbens PW, 'Synthetic-waypoint guidance algorithm for following a desired flight trajectory', Journal of Guidance, Control, and Dynamics, 33 601-606 (2010)

A path-following aircraft guidance algorithm is developed that pursues synthetic waypoints using only a small set of guidance parameters, extending the virtual-target concept to c... [more]

A path-following aircraft guidance algorithm is developed that pursues synthetic waypoints using only a small set of guidance parameters, extending the virtual-target concept to complete aircraft guidance. The operation of the synthetic waypoint guidance (SWG) algorithm is based on tracking the synthetic waypoint that travels along the designated flight path. The waypoint is considered synthetic, as its position on the flight path is a projection of the position at which the aircraft intends to be within a specified time horizon. The speed of the synthetic waypoint gradually increases, as the synthetic waypoint moves that forces the aircraft to change its heading with appropriate control to maintain a direct pursuit. The synthetic waypoint can only travel along the flight path between designated inertial waypoints, which ensure that the synthetic waypoint's movement is along the defined flight path. The guidance block within the loop executes the SWG algorithm, receiving vehicle states from the aircraft dynamics to generate guidance commands that are also fed into the control and navigation system.

DOI 10.2514/1.46204
Citations Scopus - 38
2008 Stone RH, Anderson P, Hutchison C, Tsai A, Gibbens P, Wong KC, 'Flight testing of the T-wing tail-sitter unmanned air vehicle', Journal of Aircraft, 45 673-685 (2008)

Since October 2005, the T-wing tail-sitter unmanned air vehicle has undergone an extensive program of flight tests, resulting in a total of more than 50 flights, many under autono... [more]

Since October 2005, the T-wing tail-sitter unmanned air vehicle has undergone an extensive program of flight tests, resulting in a total of more than 50 flights, many under autonomous control from takeoff to landing. Starting in August 2006, free flights with conversion between vertical and horizontal flight modes have also been undertaken. Although the latter flights have required some guidance-level ground-pilot input, significant portions of them were performed in autonomous mode, including the transitions between horizontal and vertical flight. This paper considers the overall control architecture of the vehicle, including the different control modes that the vehicle was flown under during the recent series of tests. Although the individual controllers for each flight mode are unremarkable in themselves, it is notable that the aggregate system allows the vehicle to fly throughout its entire flight envelope, which is considerably broader than that of conventional fixed- or rotary-wing vehicles. The performance of the controllers for the different flight modes will also be considered, with a particular focus on hover dispersion results, in differing wind conditions. The majority of these flights were performed on a tether test rig during autonomous control development, to ensure vehicle safety with minimal impact on vehicle dynamics. The demonstration of autonomous flight under the constraints imposed by the tether system in winds up to 18 kt is a significant achievement. Results from the more recent horizontal flight tests with conversions between vertical and horizontal flight are also presented. Most important, these results confirm the basic feasibility of tail-sitter vehicles that use control surfaces submerged in propeller wash for vertical flight control. Copyright © 2007 by University of Sydney.

DOI 10.2514/1.32750
Citations Scopus - 94
2000 Grocholsky B, Durrant-Whyte H, Gibbens P, 'Information-theoretic approach to decentralized control of multiple autonomous flight vehicles', Proceedings of SPIE - The International Society for Optical Engineering, 4196 348-359 (2000)

Decentralized systems require no central controller or center where information is fused or commands generated. Information theoretic ideas have been previously used to develop op... [more]

Decentralized systems require no central controller or center where information is fused or commands generated. Information theoretic ideas have been previously used to develop optimal fusion algorithms for decentralized sensing and data fusion systems. The work described in this paper aims to develop equivalent algorithms for the control of decentralized systems. The methods and algorithms described center on the use of mutual information gain as a measure in choosing control actions. Two example problems are described; area coverage for purposes of surveillance and navigation, and sensor management for cuing and hand-off operations. The motivation for this work is the control of multiple unmanned air vehicles (UAVs). © 2000 SPIE.

DOI 10.1117/12.403734
Citations Scopus - 29
2000 Sukkarieh S, Gibbens P, Grocholsky B, Willis K, Durrant-Whyte HF, 'A low-cost, redundant inertial measurement unit for unmanned air vehicles', International Journal of Robotics Research, 19 1089-1103 (2000)

This paper discusses the development of a low-cost, redundant, strapdown inertial measurement unit (IMU). The unit comprises four ceramic vibrating structure gyroscopes and four Q... [more]

This paper discusses the development of a low-cost, redundant, strapdown inertial measurement unit (IMU). The unit comprises four ceramic vibrating structure gyroscopes and four QLC 400 accelerometers configured on a truncated tetrahedron design. This design allows for the optimal configuration of the eight sensors, which in turn provides for a theoretical 33% increase in information. The redundant sensor configuration also allows for fault detection, which is required for many autonomous applications. This initiative is a precursor for future developments with more sensors to provide fault isolation. The paper will also present a navigation system implementing the redundant IMU with the global positioning system. Results are provided of this navigation system being implemented in an unmanned air vehicle. © 2000 Sage Publications, Inc.

DOI 10.1177/02783640022067995
Citations Scopus - 60
2000 Magrabi SM, Gibbens PW, 'Decentralized fault detection and diagnosis in navigation systems for unmanned aerial vehicles', Record - IEEE PLANS, Position Location and Navigation Symposium, 363-370 (2000)

Autonomous Unmanned Aerial Vehicles (UAVs) are a recent technological phenomenon sweeping the world stage. Full autonomy implies that the guidance and navigation system employed m... [more]

Autonomous Unmanned Aerial Vehicles (UAVs) are a recent technological phenomenon sweeping the world stage. Full autonomy implies that the guidance and navigation system employed must exhibit the highest level of integrity. This paper looks at the parity space Fault Detection and Diagnosis (FDD) methods its applicability in fully autonomous guidance and navigation systems in a decentralized system architecture. Using the existing work as a starting point this paper identifies the effectiveness of these methods when applied to situations where both the hardware and analytical redundancy exist. One of the most important issues in FDD in navigation systems using redundant sensors relates to the integrity of the solution processing architecture. Recently this has motivated the development of multiple FDD solutions running on numerous separate processors in a decentralized computing network. Typical solutions to this problem are based on decentralized or multiple Kalman Filters running in parallel. This paper will address the use and merits of the Information Filter form of the Kalman Filter in a fully decentralized FDD framework.

Citations Scopus - 18
1995 SCHWARTZ CA, GIBBENS PW, FU MY, 'ACHIEVING VECTOR RELATIVE DEGREE FOR NONLINEAR-SYSTEMS WITH PARAMETRIC UNCERTAINTIES', INTERNATIONAL JOURNAL OF ROBUST AND NONLINEAR CONTROL, 5 139-151 (1995)
DOI 10.1002/rnc.4590050205
Citations Scopus - 2Web of Science - 1
Co-authors Minyue Fu
1993 GIBBENS PW, SCHWARTZ CA, FU MY, 'ACHIEVING DIAGONAL INTERACTOR MATRIX FOR MULTIVARIABLE LINEAR-SYSTEMS WITH UNCERTAIN PARAMETERS', AUTOMATICA, 29 1547-1550 (1993)
DOI 10.1016/0005-1098(93)90019-P
Citations Scopus - 7Web of Science - 6
Co-authors Minyue Fu
Show 14 more journal articles

Conference (22 outputs)

Year Citation Altmetrics Link
2018 Volkova A, Gibbens PW, 'Automated WAMI system calibration procedure based on multi-scale fusion and adaptive data association for geo-coding error correction', Proceedings of SPIE - The International Society for Optical Engineering (2018)

© 2018 SPIE. Wide Area Motion Imagery (WAMI) systems used on surveillance aircraft may suffer from system calibration errors associated with frequent re-installation. These geo-co... [more]

© 2018 SPIE. Wide Area Motion Imagery (WAMI) systems used on surveillance aircraft may suffer from system calibration errors associated with frequent re-installation. These geo-coding errors corrupt the quality of mapping of the tracked objects, 'movers', from the image frame into world reference frame. In this study, an automated system for calibration of the imagery captured with six-camera WAMI array has been developed. The automatic calibration was achieved by a system of several multi-scale feature classifiers adaptively applied to an image captured by the camera array dependent on the feature availability and classifier accuracy. The feature extraction and association modules were designed to be operating interchangeably on a frame from any given camera. The choice of the module was performed automatically using a decision tree designed as a part of the system architecture. Calculation of the per-frame corrections to mitigate the localisation error of the movers was performed by associating the features detected in each individual camera and features extracted from available satellite imagery used as a datum. The effects of the distance to the feature and the choice of the feature extraction module on the mover localisation accuracy have been evaluated on 300 frames (6 images each) captured with the WAMI array. Significant reduction in the magnitude of the geo-coding error (from 15.77-36.54 m to 5.42-8.55 m on average) was achieved and can be seen in improved alignment of the features projected into the frame as well as the reliable mapping of the mover trajectories across frames. Unlike similar systems, focusing on post-processing, the WAMI calibration system presented in the paper was designed for continuous parameter estimation in real-time.

DOI 10.1117/12.2304680
2017 Volkova A, Gibbens PW, 'Towards Automated Feature-Based Calibration of Wide-Area Motion Imagery', DICTA 2017 - 2017 International Conference on Digital Image Computing: Techniques and Applications (2017)

© 2017 IEEE. This study proposes a method for analysis and calibration of the geo-coding error in mapping of the moving targets, 'movers', detected in the Wide Area Moti... [more]

© 2017 IEEE. This study proposes a method for analysis and calibration of the geo-coding error in mapping of the moving targets, 'movers', detected in the Wide Area Motion Imagery (WAMI). To estimate the accurate position of the movers, an iterative Ordinary Least Squares (OLS) optimisation of the camera model parameters is performed. The OLS is set to minimise the Euclidean distance between the locations of the features projected from the camera frame and control points geo- referenced using Google Earth. A procedure for automated generation of visual features for camera calibration is proposed, alongside the manual procedure used for quality assessment of the results. A sequence of camera transformations used to map the features from the image frame is detailed as well as the iterative refinement to produce more accurate mapping. It is shown that iterative optimisation of the parameters underpinning the camera transformations from image to local navigational reference frames significantly improves the mapping accuracy, decreasing the position error from 58-100 m to 1-12 m for a set of reference points. The work also discusses the potential to apply developed camera calibration algorithms to each camera in the array, to correct the trajectories of the movers in the detected traffic flow across all frames.

DOI 10.1109/DICTA.2017.8227497
Citations Scopus - 2
2016 Volkova A, Gibbens PW, 'Extended Robust Feature-Based Visual Navigation System for UAVs', 2016 International Conference on Digital Image Computing: Techniques and Applications, DICTA 2016 (2016)

© 2016 IEEE. The current study seeks to advance in the direction of building a robust feature-based passive visual navigation system utilising unique characteristics of the featur... [more]

© 2016 IEEE. The current study seeks to advance in the direction of building a robust feature-based passive visual navigation system utilising unique characteristics of the features present in an image to obtain position of the aircraft. This is done by extracting, prioritising and associating such features as road centrelines, road intersections and using other natural landmarks as a context. It is shown that extending the system with complimentary feature matching blocks and choice of the features prevalent in the scene improves the performance of the navigation system. The work also discusses the constraints, cost and the impact of the introduced optical flow component. The algorithm is evaluated on a simulated dataset containing satellite imagery.

DOI 10.1109/DICTA.2016.7797040
Citations Scopus - 2
2016 Kuether DJ, Morrell BJ, Chamitoff GE, Bishop M, Mortari D, Gibbens PW, Coen M, 'Cohesive autonomous navigation system', 2016 AIAA Guidance, Navigation, and Control Conference (2016)

© 2016 American Institute of Aeronautics and Astronautics Inc, AIAA. All rights reserved. The ability for a robotic system to fully and autonomously interact with its environ- men... [more]

© 2016 American Institute of Aeronautics and Astronautics Inc, AIAA. All rights reserved. The ability for a robotic system to fully and autonomously interact with its environ- ment is key to the future of applications such as commercial package delivery services, el- derly robotic assistants, agricultural monitoring systems, natural disaster search and rescue robots, civil construction monitoring systems, robotic satellite servicing, and many more. An architecture that is conducive to Simultaneous Localization And Mapping (SLAM), path planning, and mission planning is a critical element of a system to be robust enough to handle such applications with true autonomy. In this paper we present an architecture that lends itself to such cohesive operation of all the aforementioned goals through the implementation of a common core database to represent the environment. We present the overall architecture followed by a description of the components of the architecture and how they interact, including: a demonstration of image processing techniques using geographic information science (GIS) analytical methods and ellipsoid feature models, an explanation of database management tools using k-vector, an outline of the SLAM approach, and a description of the path planning algorithm employed.

Citations Scopus - 1
2016 Morrell BJ, Chamito GE, Kuetherz DJ, Coenx M, Gibbens PW, 'Integration of 3D SLAM, rigid body landmarks and 3D path planning', AIAA Space and Astronautics Forum and Exposition, SPACE 2016 (2016)

© 2016, American Institute of Aeronautics and Astronautics Inc, AIAA. All rights reserved. For highly capable robots to perform the challenging tasks of satellite inspection and m... [more]

© 2016, American Institute of Aeronautics and Astronautics Inc, AIAA. All rights reserved. For highly capable robots to perform the challenging tasks of satellite inspection and maintenance, space mining, space manufacturing and planetary exploration, there is a need for strong autonomous navigation control systems that can localize, map and plan for 6 Degree of Freedom (DoF) motion. A cohesive autonomous navigation system is proposed that efficiently combines the three tasks through the use of 3D, rigid body landmarks stored in a central database. A 3D Simultaneous Localization and Mapping (SLAM) algorithm uses these landmarks to determine the 6 DoF state of the robot, while at the same time generating a map that can be directly used by a trajectory optimization algorithm. This paper analyses different methods of generating and representing the 3D rigid body land-marks and how they integrate into the navigation system. The analysis includes testing of methods on example RGB-D (color and depth) video.

2016 Volkova A, Gibbens PW, 'SATELLITE IMAGERY ASSISTED ROAD-BASED VISUAL NAVIGATION SYSTEM', ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences (2016)

There is a growing demand for unmanned aerial systems as autonomous surveillance, exploration and remote sensing solutions. Among the key concerns for robust operation of these sy... [more]

There is a growing demand for unmanned aerial systems as autonomous surveillance, exploration and remote sensing solutions. Among the key concerns for robust operation of these systems is the need to reliably navigate the environment without reliance on global navigation satellite system (GNSS). This is of particular concern in Defence circles, but is also a major safety issue for commercial operations. In these circumstances, the aircraft needs to navigate relying only on information from on-board passive sensors such as digital cameras. An autonomous feature-based visual system presented in this work offers a novel integral approach to the modelling and registration of visual features that responds to the specific needs of the navigation system. It detects visual features from Google Earth* build a feature database. The same algorithm then detects features in an on-board cameras video stream. On one level this serves to localise the vehicle relative to the environment using Simultaneous Localisation and Mapping (SLAM). On a second level it correlates them with the database to localise the vehicle with respect to the inertial frame. The performance of the presented visual navigation system was compared using the satellite imagery from different years. Based on comparison results, an analysis of the effects of seasonal, structural and qualitative changes of the imagery source on the performance of the navigation algorithm is presented. * The algorithm is independent of the source of satellite imagery and another provider can be used.

DOI 10.5194/isprs-annals-III-1-209-2016
Citations Scopus - 6
2016 Cheema P, Luo S, Gibbens P, 'Development of a Control and Vision Interface for an AR.Drone', MATEC Web of Conferences (2016)

© 2016 Owned by the authors, published by EDP Sciences. The AR.Drone is a remote controlled quadcopter which is low cost, and readily available for consumers. Therefore it represe... [more]

© 2016 Owned by the authors, published by EDP Sciences. The AR.Drone is a remote controlled quadcopter which is low cost, and readily available for consumers. Therefore it represents a simple test-bed on which control and vision research may be conducted. However, interfacing with the AR.Drone can be a challenge for new researchers as the AR.Drone's application programming interface (API) is built on low-level, bit-wise, C instructions. Therefore, this paper will demonstrate the use of an additional layer of abstraction on the AR.Drone's API via the Robot Operating System (ROS). Using ROS, the construction of a high-level graphical user interface (GUI) will be demonstrated, with the explicit aim of assisting new researchers in developing simple control and vision algorithms to interface with the AR.Drone. The GUI, formally known as the Control and Vision Interface (CVI) is currently used to research and develop computer vision, simultaneous localisation and mapping (SLAM), and path planning algorithms by a number of postgraduate and undergraduate students at the school of Aeronautical, Mechanical, and Mechatronics Engineering (AMME) in The University of Sydney.

DOI 10.1051/matecconf/20165607002
2015 Volkova A, Gibbens PW, 'A comparative study of road extraction techniques from aerial imagery - A navigational perspective', APISAT 2015 - 7th Asia-Pacific International Symposium on Aerospace Technology (2015)

The problem of road extraction from satellite or aerial imagery remains one of the most complex and challenging yet crucial in image processing. The review covers road extraction ... [more]

The problem of road extraction from satellite or aerial imagery remains one of the most complex and challenging yet crucial in image processing. The review covers road extraction methods such as mathematical morphology, road tracking and modelling, texture progressive analysis, clustering, graph cuts, normalised cuts, tensor voting, artificial neural networks, support vector machines, snakes and level sets. Among criteria upon which algorithms are analysed are level of automation, ability to incorporate a priori knowledge and inclusion of road parameters (geometric, radiometric, textural, spatial) and type of output information etc. The study explores and compares the potential of individual state-of-the-art extraction algorithms and their combinations satisfying the stated requirements to facilitate future development of generic feature extraction algorithms for real-time application in visual aerial navigation systems. The survey is supported by a tabulated summary that details the operational peculiarities and output forms of each processing technique and compares their advantages and disadvantages.

Citations Scopus - 3
2010 Medagoda EDB, Gibbens PW, 'Efficient predictive flight control', ICCAS 2010 - International Conference on Control, Automation and Systems (2010)

This paper introduces an efficient predictive control algorithm applied to aircraft flight control. The algorithm is based on characterizing the predicted response of the system t... [more]

This paper introduces an efficient predictive control algorithm applied to aircraft flight control. The algorithm is based on characterizing the predicted response of the system though the use of a direct solution to the state transition matrix found via Eigen-values and Eigen-vectors. The structure and formulation of the controller is discussed in some detail, as well as its application in constrained and unconstrained systems. An overview of tuning guidelines is presented with respect to desired length of the prediction horizon, the number of prediction points and system weights. Demonstration of the proposed predictive control algorithm on a numerical aircraft simulation will be performed comparing the performance of the aircraft under various configurations. ©ICROS.

Citations Scopus - 4
2007 Tsai AC, Gibbens PW, Stone RH, 'Terminal phase visual position estimation for a tail-sitting vertical takeoff and landing UAV via a Kalman filter', Proceedings of SPIE - The International Society for Optical Engineering (2007)

Computer vision has been an active field of research for many decades; it has also become widely used for airborne applications in the last decade or two. Much airborne computer v... [more]

Computer vision has been an active field of research for many decades; it has also become widely used for airborne applications in the last decade or two. Much airborne computer vision research has focused on navigation for Unmanned Air Vehicles; this paper presents a method to estimate the full 3D position information of a UAV by integrating visual cues from one single image with data from an Inertial Measurement Unit under the Kalman Filter formulation. Previous work on visual 3D position estimation for UAV landing has been achieved by using 2 or more frames of image data with feature enriched information in the image; however raw vision state estimates are hugely suspect to image noise. This paper uses a rather conventional type of landing pad with visual features extracted for use in the Kalman filter to obtain optimal 3D position estimates. This methodology promises to provide state estimates that are better suited for guidance and control of a UAV. This also promise autonomous landing of UAVs without GPS information to be conducted. The result of this implementation tested with flight images is presented.

DOI 10.1117/12.733453
2006 Tsai AC, Gibbens PW, Hugh Stone R, 'Visual position estimation for automatic landing of a tail-sitter vertical takeoff and landing unmanned air vehicle', 13th Annual International Conference on Mechatronics and Machine Vision in Practice 2006 (2006)
2006 Tsai AC, Gibbens PW, Stone RH, 'Terminal phase vision-based target recognition and 3D pose estimation for a tail-sitter, vertical takeoff and landing unmanned air vehicle', Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (2006)

This paper presents an approach to accurately identify landing targets and obtain 3D pose estimates for vertical takeoff and landing unmanned air vehicles via computer vision meth... [more]

This paper presents an approach to accurately identify landing targets and obtain 3D pose estimates for vertical takeoff and landing unmanned air vehicles via computer vision methods. The objective of this paper is to detect and recognize a pre-known landing target and from that landing target obtain the 3D attitude information of the flight vehicle with respect to the landing target using a single image. The Hu's invariant moments' theorem is used for target identification and parallel lines of the target shape are investigated to obtain the flight vehicle orientation. Testing of the proposed methods is carried out on flight images obtained from a camera onboard a tail-sitter, vertical takeoff and landing unmanned air vehicle. © 2006 Springer-Verlag.

DOI 10.1007/11949534-67
Citations Scopus - 4
2005 Scamps A, Gibbens P, 'Inclusion of a flight training device into the QANTAS airways A330 training program', Collection of Technical Papers - AIAA Modeling and Simulation Technologies Conference 2005 (2005)

QANTAS has been conducting A330 flight training in its full flight simulator (FFS) since January 2003 located at its training facility in Melbourne, Australia. Recently the opport... [more]

QANTAS has been conducting A330 flight training in its full flight simulator (FFS) since January 2003 located at its training facility in Melbourne, Australia. Recently the opportunity to acquire a Flight Training Device arose. The location of such a device at QANTAS' Sydney training centre coupled with a removal of portions of training from the FFS to the FTD would allow for a significant reduction in training costs as many of the current aircrew transitioning to the 330 are based in Sydney. The Airbus training footprint is based around the use of a Fixed Base Simulator for a portion of the endorsement. Traditionally a FFS has been used to accomplish this training in the absence of a dedicated FBS. This paper outlines the processes undertaken to analyze this training to determine it's suitability for training in the FTD. A this stage a 30% transfer has been identified. Much of the (training will transfer readily. Some training has been the subject of debate. Special techniques are required to be used by instructors to train these and the outcomes have been positive. Copyright © 2005 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

2005 Scamps A, Gibbens P, 'Development of a Flight Simulator Evaluation Course at QANTAS', Collection of Technical Papers - AIAA Modeling and Simulation Technologies Conference 2005 (2005)

A Flight Simulator Evaluation Course has been developed at QANTAS Airways. This course is aimed at persons who wish to acquire the skills necessary for the planning, execution and... [more]

A Flight Simulator Evaluation Course has been developed at QANTAS Airways. This course is aimed at persons who wish to acquire the skills necessary for the planning, execution and reporting of re-current simulator evaluations as required by simulator regulators. The International Civil Aviation Organization's standards for qualification are used as the basis for the course. The course has been developed initially for Australian users but has been structured around the International Regulations in a modular fashion to support international customers. The course has been scheduled to run for 5 days and includes 12 hours in full flight simulators. At the end of the course, the student will be able to evaluate simulators as either a regulatory officer or as a training centre employee under delegation. This paper outlines the development and structure of the course as it has developed. The course will be delivered in the third quarter of 2005. Copyright © 2005 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

Citations Scopus - 2
2003 Scamps A, Gibbens P, 'Development of a variable stability flight simulator as a research/education tool', AIAA Modeling and Simulation Technologies Conference and Exhibit (2003)

A teaching and research simulator has been developed at the University of Sydney. The system is based on an ex-link 707 flight simulator cabin with a three degree of freedom (3 DO... [more]

A teaching and research simulator has been developed at the University of Sydney. The system is based on an ex-link 707 flight simulator cabin with a three degree of freedom (3 DOF) motion base. The flight controls are hydraulically loaded with electric trim provided on the elevator circuit. The hydraulics are controlled by a Moog servoamplifier system which in turn receives commands form a real time digital control system. The supervisory, motion and control loading software has been developed under the Matlab/Simulink XPC Target Real Time Workshop environment which allows online modification of system parameters in real-time. A Variable Stability Module allows stability parameters to modified online. © 2003 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.

Citations Scopus - 1
2000 Gibbens PW, Dissanayake GMWM, Durrant-Whyte HF, 'A closed form solution to the single degree of freedom simultaneous localisation and map building (SLAM) problem', Proceedings of the IEEE Conference on Decision and Control (2000)

This paper presents a closed form solution to the estimation-theoretic simultaneous localisation and map building (SLAM) problem. The solution is obtained by explicit solution of ... [more]

This paper presents a closed form solution to the estimation-theoretic simultaneous localisation and map building (SLAM) problem. The solution is obtained by explicit solution of the differential Riccati equation associated with then n-landmark SLAM problem. The solution describes and explains the many experimental and theoretical results obtained so far in the study of the SLAM problem. Further, the solution, for the first time, allows a precise means of analysing the performance of different SLAM algorithms and enables the design of efficient SLAM systems.

Citations Scopus - 35
2000 Nettleton EW, Gibbens PW, Durrant-Whyte HF, 'Closed form solutions to the multiple platform simultaneous localization and map building (SLAM) problem', Proceedings of SPIE - The International Society for Optical Engineering (2000)

This paper presents a closed form solution to the multiple platform simultaneous localization and map building (SLAM) problem. Closed form solutions are presented in both state sp... [more]

This paper presents a closed form solution to the multiple platform simultaneous localization and map building (SLAM) problem. Closed form solutions are presented in both state space and information based forms. A key conclusion of this paper is that the information-state based form offers many advantages over the state space formulation in allowing the SLAM algorithm to be decentralized across multiple platforms. The benefits of operating SLAM in an information form are numerous. The additive properties of the information update make it especially attractive, as does the ability to predict estimates through any direction in time. However, of paramount importance is the well-known ability to decentralize the information filter. A general form of the continuous time inverse covariance matrix for the SLAM problem is presented to determine such properties as the initial and steady state conditions. These properties are investigated to determine their dependence and relationship to both the observation and process noise. Examination of the structure of the general form of the inverse covariance matrix also gives an insight into what information should be communicated between platforms in the decentralized architecture and how it can be managed.

Citations Scopus - 21
2000 Nettleton EW, Durrant-Whyte HF, Gibbens PW, Goktogan AH, 'Multiple platform localization and map building', Proceedings of SPIE - The International Society for Optical Engineering (2000)

This paper presents current work on decentralized data fusion (DDF) applied to multiple unmanned aerial vehicles. The benefits of decentralizing algorithms, particularly in this f... [more]

This paper presents current work on decentralized data fusion (DDF) applied to multiple unmanned aerial vehicles. The benefits of decentralizing algorithms, particularly in this field, are enormous. At a mission level, multiple aircraft may fly together sharing information with one another in order to produce more accurate and coherent estimates, and hence increase the their chances of success. At the single platform level, algorithms may be decentralized throughout the airframe reducing the probability of catastrophic failure by eliminating the dependency on a particular central processing facility. To this end, a complex simulator has been developed to test and evaluate decentralized picture compilation, platform localization and simultaneous localization and map building (SLAM) algorithms which are to be implemented on multiple airborne vehicles. This simulator is both comprehensive and modular, enabling multiple platforms carrying multiple distributed sensors to be modelled and interchanged easily. The map building and navigation algorithms interface with both the simulator and the real airframe in exactly the same way in order to evaluate the actual flight code as comprehensively as possible. Logged flight data can also be played back through the simulator to the navigation routines instead of simulated sensors. This paper presents the structure of both the simulator and the algorithms that have been developed. An example of decentralized map building is included, and future work in decentralized navigation and SLAM systems is discussed.

DOI 10.1117/12.403733
Citations Scopus - 35
2000 Magrabi SM, Gibbens PW, 'Decentralized fault detection and diagnosis using combined parity space and filter innovations based methods', National Aerospace and Electronics Conference, Proceedings of the IEEE (2000)

A decentralized system architecture is utilized through an information filter implementation of the Kalman filter. This system estimates the states pertinent in the operation of a... [more]

A decentralized system architecture is utilized through an information filter implementation of the Kalman filter. This system estimates the states pertinent in the operation of an unmanned aerial vehicle. The decentralized data fusion of an inertial measurement unit (IMU) with data from the GPS and an air data system is presented to perform fault detection and diagnosis.

1991 GIBBENS PW, SCHWARTZ CA, FU MY, 'DYNAMIC DECOUPLING OF MIMO SYSTEMS - NONLINEAR CASE', PROCEEDINGS OF THE 30TH IEEE CONFERENCE ON DECISION AND CONTROL, VOLS 1-3, BRIGHTON, ENGLAND (1991)
Co-authors Minyue Fu
1991 GIBBENS PW, FU MY, 'OUTPUT-FEEDBACK CONTROL FOR OUTPUT TRACKING OF NONLINEAR UNCERTAIN SYSTEMS', PROCEEDINGS OF THE 30TH IEEE CONFERENCE ON DECISION AND CONTROL, VOLS 1-3, BRIGHTON, ENGLAND (1991)
Citations Scopus - 5
Co-authors Minyue Fu
1991 Gibbens PW, Schwartz C, Fu M, 'Dynamic decoupling of MIMO systems: Linear case', Proceedings of the American Control Conference (1991)

For systems which do not have a diagonal interactor, a bicausal precompensation or static state feedback is insufficient for decoupling. Conditions are characterized under which a... [more]

For systems which do not have a diagonal interactor, a bicausal precompensation or static state feedback is insufficient for decoupling. Conditions are characterized under which a multi-input, multi-output (MIMO) system can be made decouplable by a diagonal dynamic precompensation. More specifically, the authors determine necessary and sufficient conditions under which diagonal dynamic precompensation exists which achieves a diagonal interactor.

Citations Scopus - 1
Co-authors Minyue Fu
Show 19 more conferences
Edit

Research Supervision

Number of supervisions

Completed11
Current0

Past Supervision

Year Level of Study Research Title Program Supervisor Type
2018 PhD Robust Modular Feature-Baased Terrain-Aided Visual Navigation and Mapping
<p>Abstract</p><p>The visual feature-based Terrain-Aided Navigation (TAN) system presented in this thesis addresses the<br />problem of constraining the inertial drift introduced into the location estimate of Unmanned Aerial Vehicles<br />(UAVs) in a GPS-denied environment. The presented TAN system utilises salient visual features representing<br />semantic or human-interpretable objects from on-board aerial imagery and associates them to a database of<br />reference features created a-priori, through application of the same feature detection algorithms to satellite<br />imagery. Correlation of the detected features with the reference features via a series of the robust data association<br />steps allows a localisation solution to be achieved within a nite absolute bound precision dened by the certainty<br />of the reference dataset.<br />The capability of a UAV to <br />y and navigate autonomously is critical to the success of long-range surveillance<br />and mapping missions. During these missions a UAV is subject to the risk of losing its primary source of<br />navigation information. The potential loss of the regular position update forces the system to rely on inertial<br />sensors on-board, which are subject to signicant position drift, also known as 'inertial drift'. This rapidly<br />growing localisation error compromises the overall usefulness of the platform, leading to inaccuracies in the<br />performed mission or inability to complete it.<br />The fusion of additional sources of localisation information has been a preferred solution to inertial drift<br />problem, including the integration of such sensors as altimeter, radar, and most importantly - optical cameras.<br />Optical cameras unlike radar, are passive meaning that they do not emit the signal in order to sense the<br />environment, which is of particular interest to Defence circles. The state-of-the-art Visual Odometry (VO)<br />approaches utilise the on-board cameras to estimate the motion of the platform to constrain the inertial sensor<br />drift. Although this approach solves the motion estimation problem it does not provide an absolute position<br />update that is critical to accurate localisation and navigation of the platform. Correlating the salient features<br />detected in the imagery to the features dened in the inertial space, as demonstrated in this thesis, facilitates<br />an absolute position update removing the quadratic drift of inertial sensors or linear drift of VO solutions.<br />The idea of deriving a localisation update from the aerial imagery comes from visual <br />ight procedures followed<br />by pilots as dened by Visual Flight Rules (VFR). When a pilot identies a unique landmark or a unique area,<br />they can use it to locate themselves on the map and understand the relative aircraft orientation with respect to<br />the landmark. Similarly, the system presented in this thesis utilises visual features identiable by a human pilot,<br />the so-called semantic features, to automatically derive an absolute aircraft position and orientation update.<br />The Visual Navigation System (VNS) presented identies visual features in the aerial imagery, adaptively selects<br />them, and associates them with a pre-built map using minimal mathematical feature descriptors. The position<br />and orientation update resulting from the Data Association (DA) step is fused into the Extended Kalman Filter<br />(EKF) lter. A unique logic layer, optimised for the reliability of the EKF update selects the best features from<br />several feature threads running on the imagery in parallel and produces a reliable localisation update across a<br />wide range of environments.<br />The feature-based VNS presented in this thesis was initially developed for a navigation application. Extensive<br />testing on simulated imagery generated using multi-year satellite image datasets has led to the development of<br />a series of unique feature extraction and association techniques. The demonstration of the performance of<br />the system attracted interest from Defence groups, which led to the extension of the system application into<br />the mapping domain. This in turn has been based on the real (not simulated) <br />ight data and imagery. In<br />the mapping study the full potential of the system, being a versatile tool for enhancing the accuracy of the<br />information derived from the aerial imagery has been demonstrated. Not only have the visual features, such as<br />road networks, shorelines and water bodies, been used to obtain a position 'x', they have also been used in<br />reverse for accurate mapping of vehicles detected on the roads into an inertial space with improved precision.<br />Combined correction of the geo-coding errors and improved aircraft localisation formed a robust solution to the<br />defence mapping application.<br />A system of the proposed design will provide a complete independent navigation solution to an autonomous<br />UAV and additionally give it automatic object tracking capability.</p>
Aerospace Engineering, University of Sydney Principal Supervisor
2018 PhD Enhancing 3D Autonomous Navigation Through Obstacle Fields - Homogeneous Localisation and Mapping, with Obstacle-Aware Trajectory Optimisation
<p>ABSTRACT</p><p>The capability of small flying robots, such as quadrotors and free-floating satellites, are making<br />them useful tools for a wide range of applications. Quadrotors can be used for search and<br />rescue, facility inspection, infrastructure surveying and parcel delivery. Free-flying satellites<br />can be assistants inside space stations and monitor the outside of space stations and satellites. A<br />critical capability to enable these applications is autonomous navigation near obstacles. Autonomous<br />navigation is a challenge for small flying robots as they have limited payload capacity, hence require<br />low-powered, low-weight sensors, and efficient computation. The robots also need to localise, map, and<br />plan trajectories in 3D, a significantly enhanced challenge over 2D applications. Being flying vehicles,<br />the dynamic-feasibility of planned trajectories and the control algorithms to track the trajectories are<br />also essential considerations.<br />The state-of-the-art for autonomous navigation systems is heterogeneous, with a combination of<br />many different algorithms. What is proposed here is a more homogeneous system, with the aim for<br />enhanced efficiency.<br />To determine the location of a robot, visual Simultaneous Localisation and Mapping (SLAM)<br />algorithms using stereo or depth cameras are the leading approach for small flying robots. While SLAM<br />algorithms produce a map, it is purely for localisation, so a separate 3D mapping algorithm is required:<br />producing occupancy grids or signed distance fields for trajectory planning. No algorithms can combine<br />SLAM and 3D mapping into one algorithm without the use of lidar. This work proposes the use of 3D<br />objects, modelled with Non-Uniform Rational B-Spline (NURBS) surfaces, to serve both as features for<br />SLAM and as obstacles for trajectory planning. Modelling as objects, rather than complete environments,<br />manages the computational requirements, and using NURBS surfaces allows the resolution to be varied<br />for different tasks. The proposed approach is demonstrated on sets of simulated data, demonstrating<br />tracking errors of under 2% of the total path length, mapping errors as low as 2 cm and an appropriate<br />collision-cost profile for obstacle representation in trajectory planning.<br />Leading trajectory planning approaches are also heterogeneous, with the combination of a global<br />path planner, local trajectory optimiser and reactive obstacle avoidance. This split of algorithms can<br />provide sub-optimal trajectories, though, when being used for flight close to obstacles. Presented<br />here is the Admissible Subspace Trajectory Optimiser (ASTRO), an algorithm that provides a middle<br />ground, optimising dynamics over a large horizon with consideration for complex obstacle fields. ASTRO<br />performs polynomial optimisation with the inclusion of constraints. The constraint formulation is<br />flexible to include a wide range of obstacles, including dynamic obstacles with motion models and<br />uncertainty growth. ASTRO is shown to provide comparable computation time and success rate to the<br />state-of-the-art, through batches of simulations. Flight tests on quadrotors show that the algorithm can<br />produce trajectories that are more dynamically-feasible (easier to track) than the state-of-the-art, by<br />including obstacles directly in the optimisation.<br />Trajectory tracking control for quadrotors utilises the differential flatness transformation to link<br />position and attitude controllers. There are singularities in the transformation though, and existing<br />methods to handle the singularities can fail in different scenarios. These methods are analysed in detail<br />to highlight where failures occur, and a new, robust method is proposed. The new method is successfully<br />demonstrated in aggressive flights.<br /><br />The proposed algorithms for SLAM and trajectory planning are brought together into a complete<br />system to demonstrate the homogeneous concept. This system is compared to the state-of-the-art in a<br />novel simulation framework. The results successfully prove the concept that a single 3D representation<br />can be used for localisation, mapping and planning with lightweight sensors. The current implementation<br />of NURBSLAM is shown to be less efficient and less accurate than the state-of-the-art; however,<br />it is more robust in scenarios with sparse visual features, successfully operating in cases where other<br />visual SLAM algorithms fail, and demonstrating better recovery from errors.<br />The work presented in this thesis can be built upon to evolve the SLAM algorithm further, to be more<br />efficient and accurate. Tests can be performed in more environments and with real camera data, to aid<br />development and to further characterise where NURBSLAM provides benefits over the state-of-the-art.</p>
Aerospace Engineering, University of Sydney Co-Supervisor
2017 PhD Efficient Constrained Algebraic Model Predictive Control for Aerospace Applications
<p>Abstract</p><p>Model Predictive Control (MPC) has been of signicant research interest since it was<br />proposed several decades ago. MPC gained attention as a valuable control technique<br />due to its capability of controlling multivariable processes subject to constraints. The<br />major drawback of MPC is the computational burden associated with the large number<br />of parameters it can be optimised over. One method that improves the computational<br />efficiency of the control algorithm is Algebraic Model Predictive Control (AMPC).<br />In this method the optimal control problem can be formulated with a non-uniform<br />distribution of points in the prediction horizon. The most efficient case is one where<br />the prediction horizon is reduced down to a single prediction point, thus removing any<br />additional and redundant calculations.<br />The drawback of this prediction point reduction is loss of system information and<br />hence degraded performance while the system is constrained. This thesis extends the<br />AMPC framework, improving the constraint handling performance. This is achieved<br />by introducing additional points across the time horizon at which the constraints are<br />checked. The computational efficiency of the original method is retained by recognising<br />that the critical constraint is the peak location, the time along the prediction horizon at<br />which the peak of the response occurs, with a method developed for calculating its location.<br />By constraining this peak time location, an equivalent constrained response can<br />be achieved as though every point across the horizon length was considered, however,<br />at a signicantly reduced computational cost.<br />This thesis also develops a method of expressing the optimal AMPC law explicitly.<br />This further improves the computational efficiency of the algorithm. By combining the<br />improved constraint method with the explicit control law, stability properties of the<br />constrained, closed loop system can be guaranteed a priori. Analysis of the algorithms<br />and the computational cost is performed on several different numerical simulations<br />including an aircraft model, demonstrating the capabilities of the algorithm being used<br />in aerospace applications. Realistic eects such as model uncertainty, aerodynamic<br />parameter errors and atmospheric disturbances are considered to assess the levels of<br />robustness and stability that can be achieved with the algorithm. The extensions<br />in this thesis for the AMPC algorithm are compared with both the original AMPC<br />formulation and a baseline MPC formulation to demonstrate the equivalence, benefits<br />in terms of constrained performance as well as the improved computational efficiency<br />of the presented algorithm.</p>
Aerospace Engineering, University of Sydney Principal Supervisor
2016 PhD Edge Feature and Optical Flow Terrain Aid for GNSS-Denied Airborne Visual Navigation
<p>Abstract</p><p>Conventional autonomous aircraft navigation relies on an inertial measurement unit and Global Navigation<br />Satellite System (GNSS) to correct for time-accrued dead-reckoned drift. Inertial measurements are<br />used to accurately track fast changes, whereas GNSS signals prevent the localisation solution from<br />deteriorating over time due to inertial drift. This reliance on GNSS risks disorientation of the autonomous<br />vehicle due to GNSS failure from atmospheric events, jamming or satellite destruction. This would result<br />in an inability to navigate through GNSS-denied environments. Navigation methods which do not require<br />GNSS predominantly rely on active sensors, such as radar or laser range finders, which impose critical<br />platform restrictions due to weight, volume and power draw. Furthermore, reliance on active sensors will<br />prevent any stealthy operations of an aerial vehicle.<br />For a human pilot, vision plays a major role in the methods by which they use to navigate an aerial<br />vehicle. These navigation techniques, such as Visual Flight Rules enable a pilot to localise an aircraft<br />without reliance on GNSS or any active sensors by recognising visually distinct features. These features,<br />or landmarks, may be associated with a feature map to determine location, orientation, altitude and speed.<br />Visually distinct features may include point landmarks such as buildings, however will more generally<br />consist of curve features, such as lake or river edges, forest boundaries or roads. Furthermore, a pilot may<br />also use the shape of underlying terrain features such as hills and valleys as localisation aids.<br />This thesis outlines the development and implementation of a computer vision based autonomous<br />navigation system. This system employs the same visual navigation techniques a human pilot would,<br />using curve and surface based landmarks. Visual information is fused with inertial information in a<br />probabilistic state filter in order to minimise the dead reckoning drift from pure inertial integration. This<br />visual system is separated into two main localisation research sections; curve feature tracking and terrain<br />surface matching. Curve feature tracking allows curves such as edges to be estimated, tracked and<br />associated with known features for position updates. Terrain surface matching involves the estimation<br />of the local terrain profile. This estimate may then be associated with a known terrain profile, such as a<br />digital terrain elevation map (DTEM), for position updates.<br />The Simultaneous Localisation and Mapping (SLAM) process is a method by which newly detected<br />features may be used for navigation, even through unknown environments. The first research area of<br />this thesis involves the application of the SLAM algorithms to spline features in the full six degrees<br />of freedom aerial navigation case. Image processing techniques are outlined which allow regions of<br />differing ground type to be determined, in order to detect edge curve features. The methods by which<br />splines are used to characterise features, and the mechanics of how they are updated through multiple<br />measurements, are outlined in detail. Furthermore, this thesis shows how the range to detected features<br />may be estimated. Methods which help to improve the robustness of the feature detection algorithms, as<br />well as the estimated SLAM spline feature map and data associations, are also outlined.<br /><br />Although navigation in unknown environments is an interesting topic of research, there are few<br />real-world environments which are truly completely unknown. A number of freely available databases,<br />such as Google Earth, provide convenient sources of information which may be used to localise an<br />aircraft. This thesis demonstrates the segmentation of aerial imagery retrieved from Google Earth into<br />edge features. These features are then assembled into a data-compact database of spline features, which<br />is used for Visual Terrain Aided Navigation (VTAN), allowing accurate localisation of the aerial vehicle.<br />The second research area of this thesis involves the use of Digital Terrain Elevation Maps (DTEM).<br />Some of these maps describe the shape of most of the Earth&rsquo;s landmass, such as that produced by the<br />Shuttle Radar Topography Mission. These describe the terrain profile, such as hills, cliffs and valleys. For<br />human pilots, this information is often provided in the form of a topography map. A pilot can estimate<br />the shape of surrounding terrain by observing changes in the apparent relative speed of terrain as it moves<br />by. Comparisons may then be made between this and the topography map, helping to localise the vehicle.<br />This thesis presents a method by which movement rates of imagery in a camera frame (optical flow) can<br />be used to estimate the shape of underlying terrain. The estimated terrain contour shape is then compared<br />to a freely available DTEM, allowing localisation of the vehicle. Optical flow is simultaneously used to<br />provide visual odometry information, estimating the vehicle velocity.<br />Some visually distinct features may change over time, such as water body boundaries changing with<br />varying water height. This can cause navigation problems, as localising using these features may give<br />biased position estimates. Considering river or lake edge changes can be expected to follow the shape of<br />the terrain, a Digital Terrain Map may be used to predict this change. Should changes in water level be<br />known (such as through water storage reports) changes in edge feature location can be accounted for,<br />as demonstrated in this thesis. This improves the reliability of the presented visual navigation system.<br />Conversely, should the water level change not be known, a method of using the terrain map to estimate this<br />offset is established. This allows the possibility of aerial monitoring of remote river or lake ecosystems.<br />The outlined system is demonstrated to greatly improve navigation accuracy in unknown environments<br />using real flight test data. Terrain assistance associations with a known curve feature map are shown<br />to provide visual localisation accuracies of under 20 metres. Finally, optical flow based terrain contour<br />matching techniques result in localisation accuracies approaching those of edge feature terrain assistance,<br />without requiring the use of a pre-processed feature map.</p>
Aerospace Engineering, University of Sydney Principal Supervisor
2013 PhD Airborne Vision-Based Attitude Estimation and Localisation
<p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">Abstract</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:10.0pt;font-family:'Times New Roman',serif;"></span><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">Vision plays an integral part in a pilot 1s ability to navigation and control an aircraft. As</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">such, Visual Flight Rules have been developed around the pilot's ability to see the environment</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">outside of the cockpit, to control the attitude of the aircraft, to navigate and to avoid obstacles.</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">Automation of these processes using a vision system could greatly increase the reliability and</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">autonomy of unmanned aircraft and flight automation systems. Aircraft navigation systems &middot;</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">have a high dependence on external systems and infrastructure, such as the Global Positioning</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">System, to provide localisation information for the system. This reliance on external systems</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">prevents unmanned aircraft and flight automation systems from becoming as truly independent</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">as a human pilot.</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">This thesis investigates the development and implementation of a robust vision system which</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">fuses inertial information with visual information in a probabilistic framework to navigate the</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">aircraft. There are two main areas to the visual navigation problem investigated in this thesis.</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">The first area is the visual estimation of the attitude of the aircraft. This was investigated as the</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">attitude accuracy has a large impact on the overall navigational accuracy. The second section</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">investigated was the visual localisation of the aircraft, to constrain the inertial drift from the</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">inertial solution.</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">The horizon appearance is a strong visual indicator of the attitude of the aircraft; so techniques</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">for visual horizon attitude determination were investigated in this thesis. An image</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">processing method is developed to extract the horizon interface from a camera image. The</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">developed method is evaluated against past approaches for computational performance and accuracy.</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">A number of different attitude determination methods with varying degrees of accuracy</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">are developed to generate attitude measurements from the detected visual horizon interface. The</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">simplest method and least accurate, approximates the horizon as a straight line for a bank and</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">pitch attitude measurement. The extension of this horizon line method to a multiple camera case</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">results in a horizon plane method. This horizon plane method allows the altitude effects on the</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">attitude measurement to be removed, while the improving the overall accuracy and robustness</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">of the attitude measurement. The next development step is to remove the assumption that the</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">horizon is flat and allow the horizon to take a general profile shape. A method of simultaneously</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">mapping this horizon profile shape and using it for attitude determination is developed. Inclusion</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">of the horizon profile information allows the complete Euler angle triplet to be measured.</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">The final and most accurate horizon attitude determination method developed in this thesis is</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">a terrain-aided method. Matching the observed horizon profile to a terrain map removes the</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">induced error caused by the approximations and assumptions made by the previous methods,</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">this allows for accurate attitude determination.</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">The second area of the navigation problem investigated in this thesis is visual localisation</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">of the aircraft. Extending the terrain-aided horizon attitude determination method allows the</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">combined estimation of the aircraft attitude, position and altitude from the observed horizon</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">profile. This allows for a rough position estimation. Increased localisation accuracy is achieved</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">by a visual localisation method involving the detection of road intersections and fusing them with</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">matched positional information from a database. An image processing method is developed to</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">detect road intersections and extract their shape information. The shape of the intersection</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">is used to aid the data association process and increase the measurement information when</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">fused into the navigation filter. The cross-coupling effects of attitude estimation accuracy on</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">the navigational accuracy was also investigated to show the importance of horizon detection and</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">attitude determination in the visual navigation problem.</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">Fault detection techniques and probabilistic uncertainty measures are developed for all the</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">different attitude determination and localisation methods developed. This helps increase the</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">robustness of the final fused solution. Simulation results and flight test results for the individual</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">methods are presented to highlight the advantages and limitations of each. Results for the</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">complete navigation system are then presented.</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">The developed visual system shows comparable performance to other non-vision-based systems</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">while removing the dependence on external systems for navigation. Constraining the</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">attitude using horizon detection improves the overall visual navigational and attitude solution</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">accuracy. The vision systems developed in this thesis can help to increase the autonomy of</span></p><p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"><span style="font-size:small;font-family:Arial, Helvetica, sans-serif;">unmanned aircraft and flight automation system by providing another information source which</span></p><p><span style="font-size:small;line-height:107%;font-family:Arial, Helvetica, sans-serif;">can be fused with other navigation systems.</span></p>
Aerospace Engineering, University of Sydney Principal Supervisor
2011 PhD Efficient Predictive Guidance and Control for Aircraft Applications
&lt;p&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;Abstract&lt;/span&gt;&lt;/p&gt;&lt;p&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;Model Predictive Control (MPC), has been rigourously developed and investigated over&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;many years, and has gained widespread application into systems with slow, predictable and&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;well modelled dynamics. The main reasons for using MPC is the ability to predict future&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;responses and formulate an optimised control solution that will keep the process within the&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;performance capabilities of the system. Based on its pre-emptive nature, MPC seems well&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;suited for implementation into highly dynamic nonlinear systems such as aerospace vehicles,&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;however, it is still not widely used in this context. This is mainly due to the computational&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;burdens associated with solving nonlinear system equations and a nonlinear nite horizon&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;control optimisation problems in real-time. Furthermore, the accuracy of predictions used&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;within the optimisation is heavily dependent on the accuracy of the model, which can eff ect&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;the reliability of the control process. The research of this thesis aims to address the need&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;for improved efficiency and reliability by building upon the existing framework of MPC.&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;In this thesis, an efficient model predictive control scheme is developed for application&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;into linear and nonlinear systems to address the accuracy and computational issues&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;associated with existing MPC algorithms. The proposed algorithm, known as Algebraic&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;Model Predictive Control (AMPC), is based on characterising the predicted system response&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;through the use of a direct algebraic solution to the state transition matrix found&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;via eigenvalues and eigenvectors. Through this method, the accumulation of truncation&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;errors that would arise from standard model discretisations can be eliminated, allowing for&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;accurate and reliable predictions for long prediction horizons. Furthermore, this process&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;allows for the dimensionality of the problem to be reduced by removing the need for predictions&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;to be made at a fixed sample rate, improving the efficiency of the optimisation process&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;without losing prediction accuracy. The application of MPC to higher order systems is also&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;improved, by allowing for the individual assignment of prediction horizons for the outputs&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;being tracked, and is no longer defi ned by a single, global horizon applied equally to all&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;outputs. The result is efficient, accurate control of multiple inputs and outputs based on&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;the individual response characteristics of the system being controlled, focusing mainly on&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;its application to flight control.&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;The AMPC algorithm is applied to a range of systems including a set of second order&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;systems, an inverted pendulum on a cart and linear and nonlinear aircraft models. These&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;studies are conducted to demonstrate the application of AMPC in the control of a range of&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;systems under various control con gurations. Realistic eff ects such as model uncertainty and&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;environmental disturbances on the process are also considered to determine the inherent&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;level of robustness possessed by the controller. In addition, AMPC is compared with a&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;baseline MPC formulation to highlight the benefi ts in computational speed and accuracy of&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;the proposed method, and the flexibility possessed by the algorithm in tuning for individual&lt;/span&gt;&lt;br /&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;cases.&lt;/span&gt;&lt;/p&gt;&lt;p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;This thesis also addresses the guidance segment of automated flight through the development&lt;/span&gt;&lt;/p&gt;&lt;p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;of a Synthetic Waypoint Guidance (SWG) algorithm. The SWG algorithm is&lt;/span&gt;&lt;/p&gt;&lt;p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;based on tracking a synthetic waypoint which moves along a defined flight path ahead of&lt;/span&gt;&lt;/p&gt;&lt;p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;the aircraft. Based on the principles of proportional navigation and formation flight, the&lt;/span&gt;&lt;/p&gt;&lt;p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;algorithm assigns a set of dynamics to the waypoint, allowing it to travel along the defined&lt;/span&gt;&lt;/p&gt;&lt;p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;flight based on the speed and proximity of the pursuing aircraft. As a result, the motion&lt;/span&gt;&lt;/p&gt;&lt;p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;of the synthetic waypoint becomes intrinsically linked to the pursuing aircraft, and allows&lt;/span&gt;&lt;/p&gt;&lt;p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;the aircraft to smoothly converge upon the flight path by anticipating upcoming flight path&lt;/span&gt;&lt;/p&gt;&lt;p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;changes. This minimises the flight path divergence during discontinuous changes in the current&lt;/span&gt;&lt;/p&gt;&lt;p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;flight leg. The level of performance demanded by the aircraft by the algorithm can be&lt;/span&gt;&lt;/p&gt;&lt;p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;adjusted based on the lead time of the waypoint and is carefully assessed to determine the&lt;/span&gt;&lt;/p&gt;&lt;p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;effect that guidance parameter variations have on the stability of the closed-loop system.&lt;/span&gt;&lt;/p&gt;&lt;p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;Furthermore, the SWG algorithm is tested on a range of flight configurations demanding&lt;/span&gt;&lt;/p&gt;&lt;p style="margin-bottom:0cm;margin-bottom:.0001pt;line-height:normal;text-autospace:none;"&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;different levels of performance, including a high altitude circuit and terminal phase landing&lt;/span&gt;&lt;/p&gt;&lt;p&gt;&lt;span style="font-family:Arial, Helvetica, sans-serif;font-size:small;"&gt;approach.&lt;/span&gt;&lt;/p&gt;&lt;p&gt;&lt;/p&gt;&lt;p&gt;&lt;/p&gt;
Aerospace Engineering, University of Sydney Principal Supervisor
2007 Honours Vision Systems for Flight Guidance
&lt;p&gt;Abstract&lt;/p&gt;&lt;p&gt;Vision plays an integral part in a pilot 1s ability to navigation and control an aircraft. As&lt;br /&gt;such, Visual Flight Rules have been developed around the pilot's ability to see the environment&lt;br /&gt;outside of the cockpit, to control the attitude of the aircraft, to navigate and to avoid obstacles.&lt;br /&gt;Automation of these processes using a vision system could greatly increase the reliability and&lt;br /&gt;autonomy of unmanned aircraft and flight automation systems. Aircraft navigation systems &amp;middot;&lt;br /&gt;have a high dependence on external systems and infrastructure, such as the Global Positioning&lt;br /&gt;System, to provide localisation information for the system. This reliance on external systems&lt;br /&gt;prevents unmanned aircraft and flight automation systems from becoming as truly independent&lt;br /&gt;as a human pilot.&lt;br /&gt;This thesis investigates the development and implementation of a robust vision system which&lt;br /&gt;fuses inertial information with visual information in a probabilistic framework to navigate the&lt;br /&gt;aircraft. There are two main areas to the visual navigation problem investigated in this thesis.&lt;br /&gt;The first area is the visual estimation of the attitude of the aircraft. This was investigated as the&lt;br /&gt;attitude accuracy has a large impact on the overall navigational accuracy. The second section&lt;br /&gt;investigated was the visual localisation of the aircraft, to constrain the inertial drift from the&lt;br /&gt;inertial solution.&lt;br /&gt;The horizon appearance is a strong visual indicator of the attitude of the aircraft; so techniques&lt;br /&gt;for visual horizon attitude determination were investigated in this thesis. An image&lt;br /&gt;processing method is developed to extract the horizon interface from an camera image. The&lt;br /&gt;developed method is evaluated against past approaches for computational performance and accuracy.&lt;br /&gt;A number of different attitude determination methods with varying degrees of accuracy&lt;br /&gt;are developed to generate attitude measurements from the detected visual horizon interface. The&lt;br /&gt;simplest method and least accurate, approximates the horizon as a straight line for a bank and&lt;br /&gt;pitch attitude measurement. The extension of this horizon line method to a multiple camera case&lt;br /&gt;results in a horizon plane method. This horizon plane method allows the altitude effects on the&lt;br /&gt;attitude measurement to be removed, while the improving the overall accuracy and robustness&lt;br /&gt;of the attitude measurement. The next development step is to remove the assumption that the&lt;br /&gt;horizon is flat and allow the horizon to take a general profile shape. A method of simultaneously&lt;br /&gt;mapping this horizon profile shape and using it for attitude determination is developed. Inclusion&lt;br /&gt;of the horizon profile information allows the complete Euler angle triplet to be measured.&lt;br /&gt;The final and most accurate horizon attitude determination method developed in this thesis is&lt;br /&gt;a terrain-aided method. Matching the observed horizon profile to a terrain map removes the&lt;br /&gt;induced error caused by the approximations and assumptions made by the previous methods,&lt;br /&gt;this allows for accurate attitude determination.&lt;br /&gt;The second area of the navigation problem investigated in this thesis is visual localisation&lt;br /&gt;of the aircraft. Extending the terrain-aided horizon attitude determination method allows the&lt;br /&gt;combined estimation of the aircraft attitude, position and altitude from the observed horizon&lt;br /&gt;profile. This allows for a rough position estimation. Increased localisation accuracy is achieved&lt;br /&gt;by a visual localisation method involving the detection of road intersections and fusing them with&lt;br /&gt;matched positional information from a database. An image processing method is developed to&lt;br /&gt;detect road intersections and extract their shape information. The shape of the intersection&lt;br /&gt;is used to aid the data association process and increase the measurement information when&lt;br /&gt;fused into the navigation filter. The cross-coupling effects of attitude estimation accuracy on&lt;br /&gt;the navigational accuracy was also investigated to show the importance of horizon detection and&lt;br /&gt;attitude determination in the visual navigation problem.&lt;br /&gt;Fault detection techniques and probabilistic uncertainty measures are developed for all the&lt;br /&gt;different attitude determination and localisation methods developed. This helps increase the&lt;br /&gt;robustness of the final fused solution. Simulation results and flight test results for the individual&lt;br /&gt;methods are presented to highlight the advantages and limitations of each. Results for the&lt;br /&gt;complete navigation system are then presented.&lt;br /&gt;The developed visual system shows comparable performance to other non-vision-based systems&lt;br /&gt;while removing the dependence on external systems for navigation. Constraining the&lt;br /&gt;attitude using horizon detection improves the overall visual navigational and attitude solution&lt;br /&gt;accuracy. The vision systems developed in this thesis can help to increase the autonomy of&lt;br /&gt;unmanned aircraft and flight automation system by providing another information source which&lt;br /&gt;can be fused with other navigation systems.&lt;/p&gt;
Aerospace Engineering, University of Sydney Principal Supervisor
2004 PhD Decentralised Fault Detection and Diagnosis for Guidance and Navigation Systems
<p><span lang="EN-US">Abstract</span></p><p><span lang="EN-US">The state of the art in fault detection &amp; diagnosis, state estimation and decentralization is established as a starting point of this thesis. The theory from these fields is used to extrapolate and lay out a framework for fault detection and diagnosis in decentralised multi-sensor navigation systems onboard an Unmanned Aerial Vehicle (UAV).</span></p><p><span lang="EN-US">The thesis aims to derive a framework for the hybrid application of fault detection and diagnosis methods to decentralised multi-sensor navigation architectures, through the development of theoretical tools for the use of the Information form of the Kalman Filter, the Parity Space and frequency domain analysis towards fault detection and<span>&nbsp; </span>diagnosis.</span></p><p><span lang="EN-US">Instrumental in the fault detection and diagnosis methods employed was the understanding of Kalman Filter based state estimation and the decentralisation of these methods. Fault detection was carried out using Information Filter innovations and the Parity Space, the Information Filter being the inverse covariance form of the Kalman Filter. The two methods are analysed through a geometric and frequency domain interpretation. These methods are applied in a hybrid implementation to the decentralised architecture. Expressions for the geometric dilution of precision for sensor geometries and the parity space fault vector are derived in information theoretic terms.</span></p><p><span lang="EN-US">The Parity Space method is applied to the detection of faults on a tetrahedral Inertial Measurement Unit (IMU) and results of the simulations are provided. The UAV system is simulated as six-degrees of freedom non-linear model and decentralised through the implementation of the Decentralised Extended Information Filter to provide the platform for demonstrating the decentralised fault detection and diagnosis framework. The Direct Parity Space Method, the Temporal Parity Space Method and the Information Filter innovations validation gate were applied as a hybrid for the detection of faults in a three-sensor decentralised network, comprising an unaided-IMU, a Global Positioning System (GPS) aided IMU and an Air Data System (ADS) aided IMU. </span></p>
Aerospace Engineering, University of Sydney Principal Supervisor
2003 PhD Decentralised Architectures for Tracking and Navigation with Multiple Flight Vehicles
<p>Abstract</p><p></p><p>This thesis is concerned with the development and demonstration of decentralised data fusion<br />(DDF) algorithms in airborne applications. The decentralised architectures described<br />in this thesis require no central fusion centre and no common communications medium.<br />Sensor nodes are joined in a decentralised network without any global knowledge of the<br />network topology or other node capabilities. Each sensor node is able to form a global estimate<br />based on local sensor observations and information communicated to it by adjacent<br />sensor nodes in the network. In a DDF architecture, no sensing, processing or communication<br />component is critical to the operation of the overall system so a failure of any single<br />element results in only an incremental decrease in performance rather than catastrophic<br />system failure. As no node requires knowledge of the global network topology, the system<br />can be scaled simply by connecting new sensing nodes to the system.</p><p>The main contribution of this thesis is the development of decentralised algorithms and<br />architectures able to deal with asynchronous and intermittent communications characteristic<br />of real-world sensor nodes communicating over radio networks. The algorithms presented<br />in this thesis are developed using the information form of the Kalman filter running on a<br />variety of structured and unstructured decentralised sensor networks. Although this thesis<br />is primarily concerned with structured communication topologies, that are able to make<br />complete use of global information, sub-optimal algorithms for use in networks with dynamic<br />connectivity changes are also proposed. Exact solutions to the delayed and asequent data<br />problems are developed for the information form of the Kalman filter that enable operation<br />of DDF algorithms in a broad range of real sensor networks.</p><p>The DDF architecture is applied to the problem of tracking multiple ground targets using<br />a network of airborne sensing platforms. The objective of this problem is for the sensing<br />nodes on each platform to build a composite global picture of targets in an environment<br />using both local sensor observations and information communicated from other platforms<br />in a decentralised manner.</p><p>The second important contribution made by this thesis is in the development of algorithms<br />for multi-vehicle Decentralised Simultaneous Localisation and Mapping (D-SLAM). Closed<br />form solutions to the SLAM covariance and information matrix for simplified single and<br />multiple vehicle D-SLAM problems are presented. These show that; i) in general platform<br />to platform cross information is zero, and ii) the global map information is simply the sum<br />of the map information on each platform in the system. Using these properties a D-SLAM<br />algorithm is developed which enables multiple platforms to build a common global map in<br />a fully decentralised manner. A constant time communications algorithm is also presented<br />which ensures that the decentralised algorithm scales as the map size grows large.</p><p>This thesis also describes the implementation of these algorithms in a demanding environment<br />using multiple uninhabited airborne vehicles (UAVs). The demonstrations of decentralised<br />sensing described in this thesis are believed to be the first ever of multiple<br />cooperative UAVs.</p>
Aerospace Engineering, University of Sydney Co-Supervisor
2002 PhD An Information-Theoretic Approach to Control of Multiple Sensor Platforms
<p>Abstract</p><p>This thesis describes the development of a consistent information-theoretic basis for the<br />decentralised cooperative control of multi-sensor multi-platform systems. Robotic systems<br />composed of multiple sensors and multiple platforms hold significant potential practical<br />benefits in the autonomous execution of tasks. However, the use of multiple autonomous<br />systems requires that a mechanism be developed to allow coordination and cooperation<br />between component systems in pursuit of a common goal. Current research has resulted<br />in a diversity of approaches to this challenging problem. To date, this has not adequately<br />addressed the optimality of the collective system performance and the complexity of the<br />solution process.<br />In this thesis, it is maintained that quantitative approaches to the problem of coop-<br />eration and coordination in multiple robot systems is essential in allowing coherent and<br />extensible implementations. The approach taken builds on the established principles and<br />architecture developed for decentralised data fusion (DDF) problems. Each vehicle and<br />sensor system is considered to be a distinct decision maker within a team. Each has<br />an individual information-theoretic utility measure that captures the inter-dependencies<br />among members. Together these utilities constitute the team objective.<br />The key contributions of this thesis lie in the quantification and study of cooperative<br />control between sensors and platforms using information as a common utility measure.<br />In particular:<br />&sup2; Investigating the causes and consequences of coupling in multi-sensor problems.<br />&sup2; Demonstrating scalable coordinated control of multi-platform sensing based on local<br />decision making and decentralised data fusion.<br />&sup2; Demonstrating cooperative control among coupled sensors through anonymous com-<br />munication and negotiation.<br />The work described in this thesis provides a quantitative and analytic underpinning<br />for the future development of multi-sensor and multi platform teams.</p>
Aerospace Engineering, University of Sydney Principal Supervisor
2002 PhD Autonomous Trajectory Tracking for Aircraft Using Robust Nonlinear Leap-Ahead Single-Step Optimal Control
.
Aerospace Engineering, University of Sydney Principal Supervisor
Edit

Professor Peter Gibbens

Position

Professor
School of Engineering
Faculty of Engineering and Built Environment

Contact Details

Email peter.gibbens@newcastle.edu.au
Phone (02) 405 53212

Office

Room ES333
Building ES
Location Callaghan
University Drive
Callaghan, NSW 2308
Australia
Edit