Research On Target Hunting Strategy Based On Multi-agent Formation To Avoid Obstacles

Multi-agent system is a hot research topic in recent years.It has more advantages than single agent in dealing with complex problems,and has important applications in aviation industry,medical care,military warfare and other fields.Multi-agent cooperative control is a key enabling technology to complete cooperative tasks.Due to the complexity and uncertainty of the environment,multi-agent cooperative algorithms cannot achieve expected results.This paper studies the formation obstacle avoidance and target rounding problems in multi-agent cooperative control.The main research contents of this paper are as follows:Firstly,in order to solve the problems of formation instability and unsatisfactory obstacle avoidance,this paper proposes an adaptive multi-agent formation obstacle avoidance strategy.In the formation stage,this paper adopts the multi-objective assignment algorithm to solve the problem of target node determination and conflict,and uses the Hungarian method to solve the optimal assignment problem to quickly form the desired formation and avoid frequent changes of positions of the agents.The limitations of the artificial potential field method,modify the repulsion parameters to make the resultant force smoother,introduce virtual force to change the movement direction of the agent,and help the agent escape from the local minimum.The agent flexibly passes through the obstacle area;in the formation movement stage,in view of the instability of the formation,this paper establishes a feedback control law,the follower automatically adjusts the speed and angular velocity,and establishes the potential field between the agents,which improves the overall stability of the formation sex.The effectiveness of the algorithm is verified in simulation experiments.Secondly,in order to quickly round up the target agent,this paper proposes a multi-agent round-up strategy based on dynamic round-up points.Modeling for the round-up problem,dynamically setting round-up points around the target agent,combined with the adaptive formation obstacle avoidance strategy proposed above,using a virtual pilot for location planning,so that the round-up agent can follow the target agent,and at the same time use the negotiation The distribution method makes the round-up agents evenly distributed around the target agent to achieve round-up.Finally,multi-agent reinforcement learning is used to solve the round-up problem.In view of the shortcomings of the Multi-Agent Deep Deterministic Policy Gradient(MADDPG)algorithm,a priority experience playback mechanism is added in the sampling process to give samples different Sampling weight,through experience extraction algorithm to obtain experience samples for learning,improve learning efficiency.The effectiveness of the algorithm is verified in simulation experiments.