Research On Obstacle Avoidance Method Of Indoor Mobile Robot Based On Trajectory Prediction

With the rapid development of computer technology and sensor technology,robot technology has also made great progress,and robot has been more and more applied in people's daily life.Considering the particularity of the hospital environment,the work pressure of medical staff is large.Therefore,using robots to replace the simple work of nurses,such as patient inquiry,can reduce the work pressure and labor cost of medical staff.In the hospital environment,there are many people and other working robots.It is very challenging for mobile robots to navigate safely in dynamic environment.Therefore,on the basis of traditional obstacle avoidance methods,this thesis proposes the following obstacle avoidance process: firstly,the obstacles are extracted based on geometric features of lidar data to realize the recognition of general obstacles.Interaction multiple model(IMM)and support vector regression(SVR)are used to predict trajectory of dynamic obstacles.Then,based on the cost map,a new layer of map is built to represent the impact of future pedestrian trajectory,and the impact of this layer on path planning algorithm is verified.The specific work content is summarized as follows:Firstly,lidar data is extracted based on geometric features,and the circular model is used to replace the obstacle model to realize the recognition of obstacles.IMM Method which needs pre-determined models,is not suitable for the complex pedestrian motion.The method based on SVR does not need pre-determined model,it can learn the motion model of obstacles.In order to solve the problem that the standard SVR does not support multi output,this thesis uses the multi output SVR.Compared with the standard SVR,it can reduce the error of trajectory prediction.After getting the trajectory of the obstacle,the layered cost map in ROS is used to model the environment.On this basis,a new layer of cost map is built.At the same time,the Gauss function is used to update the cost around the dynamic obstacle.The performance of path planning algorithm is changed the form of the cost map.We have made some optimizations for dynamic window approach(DWA)algorithm and A star algorithm.Finally,in the actual office environment,the above trajectory prediction algorithm and the improved path planning algorithm are verified.It can deal with the indoor dynamic obstacle avoidance in the case of less pedestrians,and improve the ability to avoid dynamic obstacles.