Research On Multi-sensor Autonomous Recycling Navigation Technology For Mobile Robots

Nowadays,robots are used to replace the works that are operated in harsh conditions or countless mechanical processes.Moreover,robots are hired to perform some tasks that humans can not complete due to various conditions.Robot technology has found an increasingly wide utilization,at the same time,machine vision is becoming much more popular with the application and research of robots thanks to the great improvement of computational power.Among the technical study of mobile robot,the autonomy of robots is one of the current research focus,especially in an unknown environment.However,in practical application,due to the influence of the positioning accuracy,the construction of the environment map is not precise enough.The fusion of the inertial device can effectively improve the positioning effect of the robots.In this paper,we firstly model the inertial navigation system,and then establish and calibrate the error model.Next,we model the pinhole model,complete camera calibration and correct distorted images,and preprocess the images captured by the camera to reduce the calculations of the system to process images.Secondly,the image feature point extraction algorithms are researched,according to the large number of mismatches in the matching pixel point pairs,the random sample consensus(RANSAC)algorithm is introduced to remove the wrong point pairs.And then the correct matching pixel points between adjacent frame pairs are used to estimate motion,two methods are compared,one is using optical flow to track pixel points and another is points matching.Aiming at the situation that the pure visual SLAM will fail to locate under the problems of fast motion and loss of feature points,the positioning algorithm of the fusion of vision and inertial navigation is studied.The pre-integration constraint of the IMU is added to the SLAM backend optimization to improve positioning Accuracy.Finally,the measurement information of the lidar is used to construct an environment map.The greedy algorithm,Dijkstra algorithm and A * path planning algorithm are compared.Under the condition of obtaining positioning information,the planned paths of various path planning algorithms are compared through experiments.