Research On Multi-source Data Fusion And Autonomous Navigation For Autonomous Cruise On Planetary Surface

In recent years,human research on celestial bodies far away from the earth has become increasingly popular,especially Mars.For the detector,due to the distance limitation,the remote control communication delay is too long,making autonomous navigation the key to the mission of the rover.It needs to combine various sensors and related algorithms to determine the pose of the rover,and complete the map creation,and then proceed Path planning and execution of related scientific tasks.This subject is aimed at this related problem,taking the prototype of the Mars rover released by NASA as the research carrier recently,through the fusion of lidar,inertial sensors and stereo camera,the simultaneous positioning of the planet surface and the construction of the map during the cruise are completed,which is a scientific experiment for the rover Provide support.First,based on the Ackerman motion principle,a cruise control model is established,which can realize the speed calculation of the 6 forward drive motors and 4 rotation angle control motors of the Rover,thus ensuring the driving capability of the rover.Secondly,simultaneous localization and mapping based on multi-sensor data fusion are studied.Based on the Lego?loam algorithm,a stereo camera is further integrated to make up for the shortcomings of insufficient information of the lidar.A reasonable combination of lidar,Inertial measurement unit and stereo camera forms a complementary advantage.While achieving accurate positioning results,the segmentation and extraction of the 3D point cloud is completed.Then,the external registration of lidar and stereo camera was completed,and obstacle detection based on stereo camera was completed.Improve the existing registration algorithm based on feature points,modify the extraction method of stereo camera feature points,and increase the results of multiple matching and use clustering to remove outliers in multiple measurement results,and find the average.The final result is improved compared to the same type of method,and it is relatively easy to implement and convenient to operate.In order to further increase the safety performance of the rover,a fast obstacle detection method based on a binocular camera is designed,so that it can identify and avoid obstacles,and guarantee the real-time performance of the algorithm.Finally,the construction of the grid map designed for the characteristics of the planet's surface is completed.The raster map generated from the extracted threedimensional point cloud map of the drivable area and the obstacle point cloud map can facilitate the representation of slopes and obstacles with a certain slope on the planet surface,and can avoid the risk of hollow obstacles on the planet surface.Completed the integrated design of the ROS-based software system of the rover.The experimental verification of each functional module of the outdoor environment data set and the indoor environment rover on-board was completed.