Cooperative control of multi-agent systems: Theory and applications to multi-UAV systems

Abstract

Cooperative control of multi-agent systems (MASs) has been a significant focus of research within the control and robotics community, driven by its diverse applications in both civilian and military domains. Recent advancements in unmanned aerial vehicle (UAV) development have opened up new possibilities for cooperative control of multi-UAV systems, showing promise for deployment in various real-world scenarios, including mapping and navigation, target enclosing and surveillance, search and rescue, and cooperative payload transportation. This thesis contributes to the progression of cooperative control theory for MASs. More importantly, it bridges the gaps and limitations between MAS theory and practical multi-UAV applications by implementing the proposed cooperative control theory in real-world scenarios involving networked UAVs. The main results of this thesis are summarised as follows. Firstly, this thesis proposes a novel robust adaptive formation control framework for multi-UAV systems operating in the presence of unknown bounded exogenous disturbances. The proposed formation controller employs the $\sigma$-modification technique to preserve the ultimate boundedness property for both adaptive gains of the distributed controller and formation tracking errors despite the influence of disturbances. In addition, this formation controller is implemented in the multi-UAV system with a cable-suspended payload to perform cooperative payload transportation missions. Secondly, this thesis develops a hierarchical control framework for MASs that addresses the fixed-time formation-containment control problem and cooperative motion planning. The framework is designed to handle practical scenarios where no agents have prior environmental knowledge, and only a subgroup of agents is equipped with obstacle-detection sensors. A novel fixed-time formation-containment controller is implemented in the inner control loop, ensuring that leader agents achieve a prescribed formation and follower agents converge inside a convex hull spanned by those leaders within a prescribed time period. The outer control loop incorporates motion planning algorithms to facilitate group obstacle avoidance and choke point navigation. Experimental results demonstrate that networked agents can safely move towards their goal positions, effectively navigating around unknown obstacles and through choke points, even with obstacle detection sensors limited to a few agents. Finally, this thesis contributes to a new type of distributed dynamic output feedback cooperative control framework for multi-agent/multi-UAV systems, utilising the Negative Imaginary (NI) systems theory. This is driven by the fact that translational dynamics of UAVs can be feedback linearised into networked double integrator NI systems. Then, distributed Strictly Negative Imaginary (SNI) controllers and 'mixed' SNI plus Strictly Passive controllers are developed to achieve various cooperative control behaviours, including time-varying group formation control, formation tracking and containment control. The proposed schemes provide flexibility in controller selection, ease of implementation and tuning, and reduced overall complexity compared to existing ones. Furthermore, relying solely on output feedback reduces sensor requirements and offers an alternative solution when full-state measurements are unavailable. This new type of dynamic output feedback cooperative control framework is also extended to address an aircraft platooning problem, where distributed flight guidance and control systems are developed to assist onboard pilots in finding feasible and effective trajectories for aircraft during the descent to landing phases. In addition, a novel affine formation manoeuvre control framework based on this dynamic output feedback controller is proposed for MASs to execute various formation manoeuvring actions, including formation translation, rotation, scaling, shearing, and combinations of these moveme

Date of Award	1 Aug 2024
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Alexander Lanzon (Supervisor)