Research On Cooperative Obstacle Avoidance And Distributed Control Algorithm Of Robot Swarm

In recent years,many real-world applications,including cooperative search and rescue,have attracted significant attention to the control of swarm robotics systems.The realization of group cooperation through interactions of each individual.Swarm robotics provides fault tolerance,task parallelism and function distribution,thus swarm robotics have the characteristic which the group performance is greater than the sum of individuals.This makes swarm robotics capable of supporting complicated tasks.The strategy development remains challenging in the field of controlling swarm robotics system,e.g.the realization of group cooperation while collision avoidance.Due to frequent and massive interactions among robots,it is hard to implement a real-time control strategy in the swarm robotics system.This phenomenon further limits the development of swarm robotics application which adopts the real-time control strategy.Thus,we need a computationally tractable method for real-time implementation.At first,this paper presents a robot adaptive control model for unknown environments.Robots use sensors directly access to environmental information,and the environmental information need to be processed.Taking the processed environmental information,the goal and obstacle avoidance task as input to select control strategy.According to the strategy and constraints to complete movement decision.Next,in order to achieve the robots cluster group consistency and collaborative obstacle avoidance,this paper studies the target based obstacle avoidance algorithm which detour to avoid obstacles.Based on path prediction and target information transformation,the artificial potential field method is improved.The distance control between the robot and the obstacle and between the robot and the neighbor robot is realized in the meantime.We evaluate the effectiveness in Gazebo simulator.In the end,this paper proposes a distributed control policy which based on neighbor reward for swarm robotic system to optimize collision avoidance and self-organization in unknown environments.Our distributed control policy generates velocity reward and obstacle reward to form reward value.This reward value is used to measure robots' motion and states.With reward value,robots can determine which neighbor robot to communicate with to learn from for adjusting its motions.By this selective interaction,we cut down the cost of communication,and improve the expandability of multi-robot system.In the meantime,this control strategy can make robot optimize the obtain of environmental information through this from its neighbors.Thus,robots could better deal with the unknown environments.Comparative experiments are conducted using multiple four-winged drones in a simulator to evaluate the effectiveness of our control and obstacle avoidance policy.Results show that proposed methods can make mass of drones find efficient,collision-free ways in an unknown environment,and can adapt in different environments even change the number of drones.