Multi-agent Cooperative Tracking Formation Control Based On Neural Network Estimation

Multi-agent has a wide range of applications at present and in the future.The cooperative tracking control of multi-agent system is characterized by both the independence and autonomy of the individual and the cooperation and cooperation of the group,so it is considered as one of the hot research directions in the field of control theory and control engineering with important significance.Most of the previous studies ignored the influence of the uncertainty of system dynamics on the motion of multi-agent.In fact,the uncertainty of system dynamics not only causes the agent to deviate from its own trajectory but also breaks the formation of the group.Therefore,it is of great theoretical and practical value to study the cooperative tracking control problem of uncertain dynamic systems.In this paper,the cooperative tracking control problem of particle and nonholonomic multiagent systems is studied for the formation tracking control problem of multi-agent systems with uncertain dynamics.The dynamics of the uncertain dynamic considered in this paper can be completely unknown and the differential convergence of the uncertain dynamic is not required.Moreover,the control algorithm and adaptive updating law proposed in this paper are completely distributed and do not need to use any global information of the graph.In this paper,neural network is used to estimate the uncertain dynamics in the agent system,and then geometric expansion method is used to solve the formation tracking control problem.At the same time,the theoretical analysis,numerical simulation and Webots robot platform experiment results all verify the effectiveness of the proposed control algorithm.The main content of this paper is as follows:Firstly,the problem of cooperative tracking control for agent systems with uncertain dynamics in three dimensional space is studied.Based on the neural network estimation of the uncertain part of the dynamics,the geometric expansion method of the sphere and the Lyapunov method are used to design the control inputs along the normal direction of the sphere,the longitude direction and the tangent direction of the orbit,so as to realize the tracking of the given orbit and the expected formation motion.The stability condition of the system is given by using Lyapunov stability theory.The results of numerical simulation prove the correctness of the conclusion.Secondly,the problem of cooperative tracking control for nonholonomic agent systems with uncertain dynamics in two dimensional space is studied.The renewal rate of the dynamic uncertain part is estimated by the neural network.The trajectory geometric expansion method is combined with the backstepping adaptive design,in which the linear velocity and angular velocity of the agent are firstly designed to realize the cooperative tracking of the control law,and then the linear velocity acceleration and angular velocity acceleration components of the agent are designed to make the error between the actual linear velocity and the virtual control law converge to 0.The stability condition of the system is given by Lyapunov stability theory.The validity of the theoretical algorithm is verified by numerical simulation.Finally,simulation experiment verification based on Webots system.We use the simulation experimental platform to simulate the real 3D world and the intelligent two-wheeled robot with physical properties and add the positioning system to the two-wheeled robot to obtain the position and other information firstly.Then the control rate of the collaborative path tracking control designed above is transformed into the control rate of the left and right wheels of the two-wheeled robot.The robot experiment is agreed to verify the effectiveness of the cooperative path following control algorithm designed in the previous paper in practical application.