Adaptive Cooperative Control For A Class Of Stochastic Nonlinear Multi-agent Systems

There are many complex systems with stochastic factors in practical engineering,such as chemical process and robot systems,etc.It is significant to study the stochastic nonlinear multi-agent system.The dynamic characteristics of each agent in the system contains stochastic characteristics,which is an important part of the control theory of multi-agent system.Using It? lemma,backstepping design method,dynamic surface control technology,graph theory knowledge and approximation of radial basis function(RBF)neural network to research the problem of cooperative control of stochastic multi-agent system in this paper.The work will be carried out from the following three aspects:1.The problem of output consensus control for a class of stochastic nonlinear strict feedback multi-agent systems under directed communication topology is studied.Each follower has stochastic characteristics,and uses It? lemma to solve the problem of the classical differential method seeking differential failure for the virtual control law.Under the design framework of the backstepping method,RBF neural network technology is used to approximate the unknown function in the system and the adaptive output consensus control law is proposed.The quartic Lyapunov function is used to prove that the proposed control law guarantees that all signals in the closed-loop system are bounded in probability.Furthermore,a class of stochastic nonlinear multi-agent system with input saturation is studied.The smooth hyperbolic tangent function is introduced to approach the saturated piecewise function,and the problem that the saturation piecewise function,which difficult to the design of input signal is solved by using the mean value theorem to simplify the form of the smooth saturation function.Then,a linear state observer is constructed to estimate the state of the system,and a first-order filter is introduced to simplify the structure of the control law.In the process of designing the control law,the singularity problem can be avoided by adding additional items.Finally,the Lyapunov function is used to prove that all signals in the closed-loop system are bounded in probability under the action of the designed control law.2.A class of stochastic nonlinear multi-agent system state feedback containment control problems consisting of multiple leaders and followers with stochastic characteristics is studied.Use graph theory knowledge to transform the containment control problem with multiple leaders into the tracking control problem with only one leader.It? lemma is introduced to solve the differential problem of virtual control law.The design method of backstepping method and neural network approach strategy are combined to propose adaptive state feedback containment control law to ensure that the outputs of all followers converge to the convex hull formed by the leader's trajectory.It is proved that all signals in the closed-loop system are bounded in probability by using Lyapunov function.On the basis of the state feedback including control problem,the containment control problem of stochastic nonlinear multi-agent with unmeasurable state is studied.A state observer is constructed for each follower to estimate the unknown state in the system.Under the framework of the backstepping method,combining RBF neural network approximation technology,state observer and graph theory tools,an adaptive output feedback containment control law is proposed.The quartic Lyapunov function is used to prove that the proposed control law guarantees that all signals in the closed-loop system are bounded in probability.3.The output feedback containment control problem of a class of stochastic nonlinear multi-agent systems is studied.For the unknown state of the system,the state observer is designed to obtain the estimation of state.Then the dynamic surface control technology is used to improve the design strategy of backstepping.In other words,a first-order filter is introduced into each agent's subsystem,which function is to filter the virtual control law and simplify the structure of the control law.Also,it can avoid the problem of "computational inflation" caused by the traditional backstepping method.In the design process,the RBF neural network is used to deal with the unknown nonlinear function in the system.In addition,in order to reduce the resource occupation of the network channel,a fixed threshold event-triggered control law is designed to reduce the number of data transmission bits in the network channel.An adaptive control law based on the input event-triggered mechanism is designed,and the influence of filter error is eliminated in the proposed control law.By means of Lyapunov function,it is proved that all signals in the closed-loop system are bounded by probability.An adaptive control law based on input event-triggered mechanism is designed,and the compensator is constructed to eliminate the influence of the filter.The Lyapunov function is used to prove that all signals in the closed-loop system are bounded in probability.