Research On Cooperative Maneuvering Problem Of Uncertain Multi-agent Systems

Multi-agent systems consist of a group of agents with the ability of perception,communication,calculation and execution.Multi-agent systems can be competent for complex tasks that individuals are difficult to complete,and have high robustness,fault tolerance and scalability.In recent years,cooperative control has shown its widespread applications,such as smart grids,unmanned ariel vehicles,and marine vehicles.In cooperative control,cooperative maneuvering is one of the hot topics.In this thesis,cooperative maneuvering of uncertain multi-agent systems is investigated,and the main contributions are listed as follows.Firstly,consensus maneuvering guided by a single parameterized path is investigated for uncertain multi-agent systems.The existing methods proposed for the consensus tracking control guided by the time-related trajectory are spatial-temporal coupling.Besides,the many existing neural-network-based direct adaptive control methods are developed based on the tracking errors,where the estimation loop and the control loop are coupled,such that the transient performance of the control signal can not be guaranteed.To address these issues,a state-feedback consensus maneuvering controller is firstly developed based on the modular design approach for the single-input and single-output uncertain multi-agent systems in the strict feedback form.The control loop and the estimation loop are decoupled by the proposed method,and the spatial-temporal decoupled control can be achieved.A neural network predictor is proposed such that the fast adaptation of the uncertain dynamics is achieved The transient analysis shows that the proposed method can improve the transient accuracy of the neural network.The accurate derivatives of virtual control laws can be obtained by a second-order linear tracking differentiator.The closed-loop system is proved to be input-to-state stable by using the cascade system stability theory.Then,an output-based neural network observer is designed based on the state reconstruction for multi-input and multi-output strict-feedback uncertain multi-agent systems,and uncertain dynamics and unkown states can be estimated simultaneously.An output-feedback consensus maneuvering controller is designed based on the proposed observer and the modified dynamic surface control method.The closed-loop system is proved to be input-to-state stable.The efficacy of the proposed state-feedback and output-feedback consensus maneuvering controllers is shown by the simulation results.Secondly,based on the above study on consensus maneuvering,containment maneuvering guided by multiple parameterized paths is investigated for uncertain multi-agent systems.Firstly,a state-feedback containment maneuvering controller is proposed based on the neural network predictor and the modified dynamic surface control method for multi-input multi-output uncertain multi-agent systems in the strict-feedback form.A path update law is designed based on the path maneuvering error feedback,and the coordination among path variables is achieved.By using the proposed method,the outputs of multi-agent systems can converge to the convex hull spanned by multiple parameterized paths.The cascade system analysis shows the input-to-state stability of the closed-loop system.Then,an output-based nonlinear neural network observer is designed for the output feedback.Uncertain dynamics and unknown states can be estimated simultaneously by using the proposed observer,and the finite-time input-to-stability of the estimation loop is established.An output-feedback containment maneuvering controller is designed based on the proposed nonlinear observer,the modified dynamic surface control method,and a super virtual leader.The closed-loop systems cascaded by the estimation loop and the control loop is proved to be input-to-state stable based on the small gain theorem and the cascade system stability theory.The effectiveness of the proposed state-feedback and output-feedback containment maneuvering controller is demonstrated by the simulation results.Finally,event-triggered containment maneuvering of multi-agent systems is investigated.In many existing time-triggered control methods,the execution of communication and actuation is periodic,such that the communication and actuation resources may be wasted.To address the above issue,an event-triggered containment maneuvering controller is developed based on the event-triggered mechanism,the neural network predictor,and the modified dynamic surface control method for a class of uncertain strict-feedback multi-agent systems subject to the constraint resource of communication and actuation.The control objective of containment maneuvering is achieved,and the resource of communication and actuation can be saved.The neural network predictor is constructed by using the aperiodic information.A third-order linear tracking differentiator is utilized to take the derivative of virtual control laws.Then,for the uncertain multi-agent systems in the presence of unknown input coefficients,a concurrent-learning-based neural network predictor is designed based on the data-driven technique to estimate the uncertain dynamics and the unknown control coefficients simultaneously,and an event-triggered containment maneuvering controller is designed based on the proposed concurrent-learning-based neural network predictor and the event-triggered mechanism.The closed-loop system is proved to be input-to-state stable,and Zeno behavior does not occur during the control process.The efficacy of the proposed event-triggered containment maneuvering controllers is demonstrated by the simulation results.