Research On Multimodal Fusion Following And Dynamic Obstacle Avoidance Of Mobile Robots

With the continuous development of industrial technology,the demand for service robots is constantly increasing.A popular trend in the design of "following" systems for following type service robots is sensor fusion,as a single sensor is difficult to adapt to complex and everchanging environments.The combination of multiple sensor fusion systems can significantly improve the reliability of obstacle avoidance following systems.This thesis adopts ultra wideband technology,LiDAR,and infrared sensing array to design and study the tracking and dynamic obstacle avoidance system of mobile robots.Among them,UltraWideBand(UWB)communication technology is used for target positioning and tracking,LiDAR is used to detect the relative position information between obstacles and robots,and infrared sensor arrays are used to detect the direction of movement of obstacles passing ahead.Propose the improved Dynamic Window Approach(DWA),improve the evaluation function evaluation index to reduce trajectory length,reduce the time to reach the target point,and improve algorithm efficiency.Design a multimodal information fusion decision-making method in the process of robot motion and obstacle avoidance.Finally,the Turtlebot2 robot was used in the ROS system to conduct tracking and obstacle avoidance experiments in different complex scenarios.The specific research content is as follows:1.A target localization and following algorithm for mobile robots was designed based on the performance characteristics of UWB.In response to the problem of dynamic obstacles during the following process,a tracking and obstacle avoidance system based on the fusion of LiDAR and UWB multi-sensor was constructed.Lidar is used for real-time detection of obstacles on the following trajectory,and UWB obtains target positioning information for target tracking.We have designed a fusion algorithm for obstacle detection using LiDAR and UWB positioning and following decision-making.The robot obtains real-time information on the relative position of obstacles through LiDAR,selects appropriate motion directions,and achieves dynamic obstacle avoidance during the following process.2.Infrared sensors have a wide range of applications in the field of robotics.In response to the inability of a single infrared sensor to obtain information beyond the distance of obstacles,this thesis designs an infrared sensor array pedestrian(or obstacle)detection system,which can not only detect distance information of moving targets but also identify the direction of target movement.When a moving target passes through an infrared array,the system detects that the infrared reflected echo signal is converted into target distance information.K-means clustering analysis is used to separate the moving target signal from background noise,and then least squares fitting is used to extract target motion direction features.Finally,a classifier is used to effectively perceive and classify eight different motion directions.3.In response to the shortcomings of the DWA algorithm that does not take into account the directionality of the trajectory to be evaluated,the smoothness of motion,and the speed of reaching the target point,and other performance indicators such as the directionality of the generated trajectory to be evaluated,the smoothness of robot motion,and the speed of reaching the target point,an obstacle evaluation subfunction and a velocity evaluation subfunction are added,while changing the weights of the obstacle evaluation function and velocity evaluation function,Improvements have been made to the DWA algorithm.The improved DWA obstacle avoidance algorithm reduces the time required for the robot to reach the endpoint,reduces the trajectory length,and improves algorithm planning efficiency.In summary,this thesis designs a tracking and obstacle avoidance system based on multimodal fusion of UWB and LiDAR,and verifies the system through robot experiments in a real environment.The experimental results show that the proposed mobile robot tracking and obstacle avoidance system based on the combination of UWB and LiDAR can achieve sTab human target tracking in complex scenarios,and can effectively avoid unknown dynamic obstacles appearing on the target tracking trajectory in real-time.In addition,the infrared sensor array can accurately detect and recognize eight directions of pedestrian movement,with an average recognition accuracy of over 0.83.The experimental results verify the feasibility of the proposed infrared sensor array based accurate detection and recognition of target movement directions.