Dr Boyang Li

This paper investigates the application of Model Predictive Control (MPC) for path tracking of a vertical takeoff and landing (VTOL) tail-sitter unmanned aerial vehicle (UAV) in hovering. In this work, the nonlinear dynamic model of a quad-rotor tail-sitter UAV including the aerodynamic effect of the wing, propellers, and slipstream was developed. The cascaded MPC controllers were then built upon linearized dynamic models. Path tracking simulations were conducted in a hardware-in-loop (HIL) environment where the UAV model and controllers were running on a PC and a flight computer independently. The simulation results show that the proposed MPC controllers are capable to perform good path tracking and the ability of disturbance rejection under limited on-board computation resource.

This paper presents a hardware-in-loop (HIL) simulation method for a tail-sitter vertical take-off and landing (VTOL) unmanned aerial vehicle (UAV). A six-degree-of-freedom (DOF) dynamic model for the tail-sitter vehicle is obtained with an aerodynamic database. An HIL simulation environment is developed that is capable of real-time dynamic simulation and supports a robot operating system (ROS)-based open-source autopilot. An independent ROS package is developed for data communication between a simulator and flight control computer. The hardware-in-loop setup is an indispensable tool for both hardware and software design of the control system for tail-sitter vehicles.

This paper investigates the multi-robot adversarial search (MuRAS) problem, which requires coordinating a team of mobile robots to search for one adversarially moving target in discrete environments. One unique challenge that MuRAS poses to the multi-robot search community in comparison to the canonical multi-robot efficient search (MuRES) problem is that the target adapts its motion model to avoid being detected by the robot team, rendering the environment non-stationary and degrading the performance of most MuRES solutions. In this paper, we first formulate MuRAS as a minimax optimization problem, i.e., the zero sum game, and then propose an algorithm, namely SMC-Searcher, a signal mediated coordination method for decentralized adversarial moving target search. SMC-Searcher enhances the canonical multi-robot search strategy by injecting a global coordination signal that prompts different and thus diversified search strategies for each robot. We demonstrate that SMC-Searcher achieves the best performance, in terms of the target's expected capture time when compared to existing multi-robot search strategies, with a simple yet illustrative example, and further compare its performance with state-of-the-art multi-robot search strategies in two canonical multi-robot search environments, namely OFFICE and MUSEUM. Additionally, SMC-Searcher is integrated into a real multi-robot system for moving target search in a self-constructed indoor environment.

This article introduces a novel robust reinforcement learning (RL) control scheme for a quadrotor unmanned aerial vehicle (QUAV) under external disturbances and model uncertainties. First, the translational and rotational motions of the QUAV are decoupled and trained separately to mitigate the computational complexity of the controller design and training process. Then, the proximal policy optimization algorithm with a dual-critic structure is proposed to address the overestimation issue and accelerate the convergence speed of RL controllers. Furthermore, a novel reward function and a robust compensator employing a switch value function are proposed to address model uncertainties and external disturbances. At last, simulation results and comparisons demonstrate the effectiveness and robustness of the proposed RL control framework.

Precision horticulture is evolving due to scalable sensor deployment and machine learning (ML) integration. These advancements boost the operational efficiency of individual farms, balancing the benefits of analytics with autonomy requirements. However, given concerns that affect wide geographic regions (e.g., climate change), there is a need to apply models that span farms. Federated learning (FL) has emerged as a potential solution. FL enables decentralized ML across different farms without sharing private data. Traditional FL assumes simple two-tier network topologies and, thus, falls short of operating on more complex networks found in real-world agricultural scenarios. Networks vary across crops and farms and encompass various sensor data modes, extending across jurisdictions. New hierarchical FL (HFL) approaches are needed for more efficient and context-sensitive model sharing, accommodating regulations across multiple jurisdictions. We present the RuralAI architecture deployment for tomato crop monitoring, featuring sensor field units for soil, crop, and weather data collection. HFL with personalization is used to offer localized and adaptive insights. Model management, aggregation, and transfers are facilitated via a flexible approach, enabling seamless communication between local devices, edge nodes, and the cloud.

Hydrodynamic forces are an important input value for the design, navigation and station keeping of underwater Remotely Operated Vehicles (ROVs). The experiment investigated the forces imparted by currents (with representative real world turbulence) and waves on a commercially available ROV, namely the BlueROV2 (Blue Robotics, Torrance, USA). Three different distances of a simplified cylindrical obstacle (shading effects) were investigated in addition to the free stream cases. Eight tethers held the ROV in the middle of the 2 m water depth to minimise the influence of the support structure without completely restricting the degrees of freedom (DoF). Each tether was equipped with a load cell and small motions and rotations were documented with an underwater video motion capture system. The paper describes the experimental set-up, input values (current speed and wave definitions) and initial processing of the data. In addition to the raw data, a processed dataset is provided, which includes forces in all three main coordinate directions for each mounting point synchronised with the 6DoF results and the free surface elevations. The provided dataset can be used as a validation experiment as well as for testing and development of an algorithm for position control of comparable ROVs.