Research On Cooperative Control Strategy Of Mobile Robot Based On Reinforcement Learning

With the continuous development of automation technology,the material transportation system in the workshop is constantly developing in the direction of intelligence.As a new type of material transportation equipment,industrial AGV has greatly improved the efficiency of material handling in the workshop because of its high flexibility and flexibility.Efficiency has gradually become the core of workshop material transportation.However,with the continuous improvement of AGV transportation capacity requirements for material transportation in the workshop,the application of a single AGV is relatively limited,and the material handling system composed of multiple AGVs can greatly expand its application scope.Therefore,this paper mainly focuses on the multi-AGV system.The main research contents are as follows:(1)Optimizing the robot trajectory tracking algorithm: The multi-robot system first needs to accurately track the global trajectory,so as to reach the designated position and execute the corresponding task.The first is the trajectory tracking controller based on the PID algorithm,but the PID algorithm is difficult to adjust the parameters,and the parameters are fixed,which cannot well adapt to the predetermined trajectory of different curvatures,so the BP network is added on the basis of the PID to make it suitable for different The curvature path outputs different PID values.Compared with the simple PID algorithm,although the tracking effect is improved,there is still a probability of tracking failure.This is due to the defect that the BP network is easy to fall into the local optimum.On the basis of the BP network,a genetic algorithm is added to optimize the network parameters globally,and a control algorithm based on the fusion of improved GA and BP-PID is designed.Accurate tracking verifies the effectiveness of the algorithm.(2)The formation control algorithm was improved: firstly,the cooperative formation control algorithm of multi-robots was analyzed,a behavior-based formation control algorithm was designed,and the formation control algorithm of row formation,column formation,V-shaped formation and diamond formation was tested.The formation simulation experiment was carried out,and secondly,it was improved to overcome the problems of formation lag and formation effect lag of the multi-machine system.(3)A multi-AGV formation algorithm based on the fusion of reinforcement learning and behavioral methods is designed: first,the relevant theoretical basis of reinforcement learning and its commonly used algorithms are analyzed,and then the deficiencies of behavioral formation control algorithms are analyzed.The application requirements of robots are getting higher and higher.When multiple AGV systems are required to form a formation,they can change according to changes in the environment to improve the flexibility and adaptability of the overall formation.Therefore,a multi-AGV formation algorithm based on the fusion of reinforcement learning and behavioral methods is designed.And three kinds of MATLAB simulation environments are designed for it,and the feasibility of the algorithm is verified by the final simulation results.(4)Finally,a physical platform of multi-mobile robots is built,the relationship between each module is analyzed,and four experimental scenarios are tested respectively.It is found that multi-AGV can realize formation in real environment.The experimental results verify the feasibility of the algorithm.