Distributed Tracking Control Of Multiagent Systems Under Complicated Constraints And Its Application

With the accelerated breakthrough and application of new generation information technologies such as wireless mobile communication and intelligent sensors,multiagent systems have attracted great attention and have been widely applied in aerospace,seabed exploration,UAV formation,smart grid and other fields.The distributed collaborative control strategy that agents complete complex tasks through local information interaction to improve the overall ability,which can not be completed by a single individual,is not only an important guarantee for high efficiency in the military and civil fields,but also the key foundation to develop artificial intelligence technology in the future.Due to the complicated system dynamics and uncertain working environment,it is necessary to consider various constraints in the control design process to provide a safety and steady operation.However,traditional constrained control methods such as ignoring the nonlinear part or local feedback linearization are inadequate to improve the safety and reliability.At the same time,the inaccurate control protocol is not sufficient to realize the full potential of multiagent systems.How to achieve the rapid and ac-curate response of multiagent systems under various constraints has become a new challenge.This thesis focuses on the distributed tracking control of multiagent systems under three types of constraints: system structure,performance index and external environment.With the help of algebraic graph theory and other related theories,this thesis studies the consensus tracking behavior of multiagent systems with single leader and multiple leaders.Considering the uncertainty caused by complicated constraints,on the basis of backstepping technique,some useful design methods such as state observer,barrier potential function,auxiliary dynamic system and adding a power integrator are applied to solve the consensus tracking problem of multiagent systems under complicated constraints,and analyse the system stability and performance.The specific research contents of this thesis are as follows:1.For multiagent systems with complicated constraints such as immeasurable state,actuator saturation and time-delay,a distributed adaptive control method based on output feedback is developed.Firstly,a nonlinear observer is established to estimate the state information directly.According to the input saturation model,an auxiliary variable is proposed to compensate the saturation.The prescribed performance constraint function is introduced to ensure that the consensus tracking error is constrained within the predesigned range.Then,under the assumption that the delayed function is bounded,an appropriate Lyapunov-Krasovskii functional is designed to compensate it.Finally,based on the stability theory,it is proved that the proposed control scheme ensures that each follower can synchronize with the leader output under the specified performance constraints.2.For a class of time-delay nonstrict-feedback multiagent systems with stochastic disturbances and full state constraints,a distributed adaptive control scheme is proposed.Firstly,by introducing the variable separation technique,the difficulties caused by the nonstrict-feedback structure of multiagent systems are solved,and thus the algebraic loop problem is avoided.Then,an adaptive mechanism is proposed to compensate the filtering error and reduce the side effect of filter errors.Furthermore,based on the stochastic stability theory and It(?) differential principle,a log type barrier Lyapunov function is designed to ensure that the full state constraints are not violated.A distributed adaptive neural network control scheme is proposed to realize the consensus tracking of multiagent systems and ensure that all closed-loop signals are bounded in probability.3.For the containment control problem of nonstrict-feedback multiagent systems under stochastic disturbances,a finite-time containment control algorithm is proposed.Firstly,a linear observer is established to obtain the agent states with simplified algorithm.Secondly,the design difficulties caused by nonstrict-feedback are subtly addressed with RBF neural network.A second-order Levant sliding mode differentiator is developed,which not only reduces the complexity of the traditional backstepping controller design,but also solves the singular problem of recursive backstepping-based finite-time controller.Meanwhile,an error compensation mechanism is designed to reduce the influence of filtering error which provides a more practical control design idea.4.For a class of nonlinear multiagent systems with time-varying mismatched disturbances and actuator fault,a smooth continuous finite-time containment control strategy is proposed.Firstly,algebraic graph theory is introduced to describe the information interaction between agents,and then a disturbance observer is constructed to counteract the adverse effects of mismatched disturbances.Then,considering the coupling problem between actuator fault and unknown disturbances,a new adaptive compensation mechanism with two adaptive law is introduced.Furthermore,the adding a power integrator strategy is applied to develop continuous finite-time controller with simplified control design process,which not only avoids the system input chattering phenomenon,but also realizes the finite-time containment control of multiagent systems with complicated constraints.Finally,based on Lyapunov stability theory,it is proved that all closed-loop signals converge to the neighborhood of the equilibrium point within finite time.5.For the nonholonomic mobile robots under safety distance constraints,optimal backstepping technique is introduced to solve the optimal containment control problem.Firstly,in order to avoid the potential collision,an auxiliary variable is constructed to predesign the ideal convergence position and ideal collision avoidance distance.Then,the constrained problem is transformed into an unconstrained tracking control problem.Combined with backstepping technique and reinforcement learning,a coordinate transformation is introduced to avoid solving the HJ equation of an augmented system.In order to simplify the learning structure,the single network structure is adopted to obtain the solution based on neural network.Secondly,according to the gradient descent method and the system stability analysis,a novel update law of the critic network is proposed to relax the constraints on the initial condition.Finally,based on Lyapunov stability theory and related experiments,it is verified that the proposed controller can ensure that the followers converge to the convex hull formed by the leaders without collision.