Research On Visual Location And Path Planning Of Mobile Robots

Autonomous localization and navigation technology of mobile robots is a hot research topic in recent years.This topic focuses on the positioning problem of mobile robots using visual sensors and the backloop detection problem used to eliminate the accumulated errors in the positioning process,and analyzes the point-to-point path planning problem of the robot and the multi-robot path planning problem.The specific contents are as follows:Firstly,aiming at the visual positioning technology of mobile robots,a visual odometry design method based on feature method is adopted.In the feature extraction and description part of the method,ORB,SURF and SIFT algorithms are used respectively.The time complexity and positioning accuracy of the visual odometry methods based on these three algorithms are compared,and the most efficient ORB feature points are selected.On this basis,the combination of EPnP algorithm and BA algorithm is used to estimate and optimize the pose of the camera.Then,in order to solve the problem that cumulative errors lead to the estimated trajectory gradually deviating from the real trajectory in the process of robot moving,a loop detection algorithm based on word bag model is used to eliminate the cumulative errors.Feature extraction and description are performed on the laboratory images collected by Kinect,and these features are trained to generate the dictionary represented by each image.Finally,the similarity between these images is described by the weight score,and the images from the same scene are detected by the algorithm,that is,the loop is detected.Next,the point-to-point path planning algorithm and the multi-robot path planning problem are studied.Aiming at the problem that the traditional A* algorithm can not guarantee the smoothness of the planned path and avoid the collision on the same path,the corner constraint and linear motion constraint are added to the algorithm,and the multi-robot path planning method based on congestion control is adopted.After that,the improved algorithm is simulated many times.The experimental results show that on the basis of not increasing the running path to a large extent,the improved A* algorithm significantly reduces the number of robot corners,and the number of robot collisions in the multi-robot system with congestion control is also reduced to a certain extent.Finally,the visual localization algorithm and loop detection algorithm are tested in the laboratory environment on the mobile robot platform.The experimental results show that the visual odometry method based on ORB features has both the lowest time complexity and high positioning accuracy,among which the average positioning error based on ORB featurepoints is 3.91%.The BoW algorithm adopted in this topic respectively creates the dictionary and calculates the similarity between the collected images.Two images from the same scene are successfully detected,which verifies the effectiveness of the word bag model algorithm.Finally,the acquired information is used to create the map of the laboratory environment.