| With the rapid development of science and technology,intelligent vehicle has been widely used in logistics distribution,disaster rescue and other fields.However,the intelligent vehicle equipped with a single sensor receives incomplete environmental information,which makes it unable to complete the tasks of positioning,mapping and navigation.In order to solve these problems,an intelligent vehicle navigation and obstacle avoidance system is designed in this paper.The system can complete the positioning and mapping of intelligent vehicle,and can realize effective navigation and obstacle avoidance in complex environment.Firstly,the Wilhelm Ackermann model of the intelligent vehicle is modeled and analyzed,and the position of the sensors relative to the intelligent vehicle is transformed into relative coordinate system.The model of each sensor is analyzed.In order to solve the problem of laser point cloud distortion in the process of lidar movement,a method of using odometer to correct the laser point cloud is proposed.In addition,the calibration experiment of RGB-D camera is carried out and the camera parameters are determined.At the same time,the principle of another kind of image sensor-infrared thermal imager is analyzed,and the function of infrared temperature measurement is realized in the program.Secondly,this research focuses on the intelligent vehicle in the multi-scene mapping problem.Based on laser SLAM,Cartographer algorithm based on graph optimization and Gmapping algorithm based on filtering are studied respectively.At the same time,the laser radar and the Infrared Thermal Imager are fused together,and when the high temperature object is detected in the raster map,the surrounding raster state is changed,change the state around the grid high temperature point to the occupied state.In addition,RTABMAP algorithm is proposed to fuse RGB-D camera and lidar.The algorithm collects the data of lidar,RGB-D camera and odometer,and stores it in the memory management mechanism node to extract the feature of the node.Through the matching times of visual terms between nodes,the weights of nodes are updated,and the discrete Bayesian filter estimation is used for loop detection to optimize the local map,and finally the global map is constructed.Then,in order to solve the location problem in intelligent vehicle navigation,this paper uses AMCL algorithm.Aiming at the problem that the global path planning algorithm A * can get the global path but can not avoid the dynamic obstacle effectively,the Teb algorithm for local path planning can avoid obstacles,but the path is only local optimal,not global optimal.This paper proposes a navigation obstacle avoidance method based on A * and Teb algorithm.Finally,a real navigation experiment is carried out to verify the feasibility of the proposed algorithm.The experiments of SLAM navigation and RGB-D imagebuilding navigation are carried out.Experimental results show that the fusion algorithm can collect more abundant environmental information,not only to avoid static obstacles,but also to avoid dynamic obstacles.In addition,the navigation experiment with high temperature proves that the lidar fusion infrared thermal imager can not only avoid the high temperature object,but also keep away from it safely to ensure the safety of the intelligent vehicle. |
