Research On Obstacle Avoidance Algorithm Of Mobile Robot In Dynamic Environment

In the past few decades,with the advancement of the technology,the moving ability and intelligence of the mobile robot have been significantly improved.And the application scenario has also changed from the original static environment to more complex dynamic environment.Thus,with more accurate ability of obstacle avoidance becomes more necessary for the robots.Recent years,the mobile robots have been applied to some dynamic environment,but the moving safety is still unstable.This means that the moving safety of the mobile robots in dynamic environments is still an unsolved problem.Therefore,the research on robot obstacle avoidance algorithm is quite necessary also with great significance no matter in theoretical research or practical application.This thesis aims at improving the algorithm of velocity obstacle and solving the obstacle avoidance problem of the robot in the application of dynamic environment.The main contents and the efforts are as following:(1)Through the analyzing of the characteristics of the dynamic environment,the characteristics of the dynamic obstacles,and the robot's motion ability,we clarified the problems need to be considered in the dynamic environment,and confirmed the research direction for improving the algorithm of velocity obstacle.(2)The disadvantage of the velocity obstacle algorithm is that it does not take the robot dynamic constraints into consideration.We came up with the idea of the hierarchical velocity obstacle algorithm which combined the advantages of both the generalized velocity obstacle algorithm and the velocity obstacle algorithm.This new algorithm set up the dynamic constraints to the robot.Through hierarchical collision detection,unnecessary calculations in the generalized velocity obstacle algorithm are effectively avoided.Through the evaluation of collision time and heuristic control,the problem of obstacle avoidance conservatism and other problems that may have no solutions been solved.By simulation,the validity of the algorithm was verified.The simulation results show that the proposed algorithm generates a more reasonable trajectory and reduces the algorithm running time by at least 15.9% in different scenarios.(3)Targeting on the uncertain problems in dynamic environment,the hierarchical velocity obstacle algorithm have been improved.The position and speed uncertainty have been controlled in the building process of velocity obstacle based on the original algorithm.An obstacle trajectory prediction model was constructed with the combination of long short term memory network and mixture density network.It's used to deal with the uncertainty of obstacle trajectory and form the probability distribution of the predicted trajectory of the obstacle,the certainty description of the obstacle prediction trajectory was replaced in the improved algorithm.The probabilistic collision detection method was proposed based on the uncertain distribution of trajectory points.Finally,through the simulation verified the effectiveness of the new algorithm on dealing the uncertainty.(4)Aiming at the the existence of the static obstacle in the dynamic environment,first,the static obstacle collision evaluation function was added for hierarchical velocity obstacle algorithm.Second,When the distribution of the static obstacle in the environment is relative complex,the global path planning should be constructed in the static map and will be used to lead the hierarchical velocity obstacle algorithm to proceed the avoidance.Finally,the simulation verification was done to show the avoidance of dynamic and static obstacle coexist under complete dynamic environment.