Design And Implementation Of Navigation System For Service Robot In Dynamic Environment

With the rapid development of robot-related technologies and the continuous expansion of the robot market,the application of service robots in various scenes such as hospitals,shopping malls,and museums is becoming increasingly widespread.However,the navigation system of service robots still faces many challenges in complex and dynamic environments.For example,if pedestrians are influenced by their mobile phones or peers during movement,they may not be able to observe the surrounding environment in a timely manner,resulting in collisions with the moving service robot.Therefore,how to efficiently avoid pedestrians while ensuring their safety and social comfort has become an urgent and important issue in the field of service robots.In response to this problem,this thesis takes indoor service robots as the research object and designs a navigation system that can combine pedestrian motion state information.The specific research content is as follows:(1)A pedestrian detection method that combines 2D Li DAR and RGB-D camera is proposed.The 2D Li DAR is used to identify the legs of pedestrians.First,the 2D point cloud is filtered,then clustered,and feature extraction is performed on the segmented point clusters.Finally,a multi-scale Adaboost classifier is used to detect human legs and determine the position of pedestrians.At the same time,an RGB-D camera is used to recognize the upper body of the person based on depth template matching.Considering that the detection range of the 2D Li DAR is wide but the accuracy is relatively low,while the detection range of the RGB-D camera is small but the accuracy is high,the two can complement each other’s advantages.This thesis uses the global nearest neighbor data association algorithm to fuse the detection results of the two,obtaining pedestrian position information.Finally,based on the Kalman filtering algorithm,the pedestrian’s motion state is predicted to obtain information on the pedestrian’s direction of motion and speed.(2)A local path planning algorithm considering pedestrian motion status information is proposed.Firstly,the relative speed between the robot and pedestrians is calculated based on acquired pedestrian positions and motion status information,and a dynamic pedestrian area is constructed based on a multi-layer cost map.Secondly,a pedestrian avoidance strategy is proposed based on the dynamic pedestrian area,and the evaluation function of the DWA algorithm is extended.Finally,the planned points of the A* global path planning algorithm are processed,and only the turning points outside the pedestrian area are retained as local target points for the DWA algorithm.The A*algorithm and the improved DWA algorithm are then combined.(3)A experimental platform for a service robot navigation system has been constructed,where the system’s various modules were tested in both simulation and real-world environments.Volunteers were invited to experience the service robot and rate its performance,and paired t-tests were conducted on the results.The analysis showed that the navigation system built in this thesis improved scores by 77%,66%,and 62% in three areas: whether the system blocked the path,the degree of interference with pedestrians,and reaction speed.These results validate the effectiveness of the proposed navigation system in this study.