Research And Implementation Of SLAM System For Mobile Robot Based On Fusion IMU

Mobile robot cannot complete the "task" quality without positioning and navigation technology.Market positioning and upgrade industries make SLAM(Simultaneous Localization and Mapping)technology become the most important breakthrough of autonomous navigation technology.But because the SLAM technology for mobile robot is affected by the difference between products and hardwarethe,the complexity between system integration and optimization,the fragmentation between algorithm and software and so on,which make the indoor SLAM still in the research stage.In this paper,in order to solve these problems,the IMU(Inertial Measurement Unit)and SLAM system were fused.The improvement and optimization methods of ORB-SLAM(simultaneous localization and map building based on ORB)system were studied in detail before and after combining IMU.The main research work is following:First of all,the SLAM schemes were selected and the system of mobile robot was built.According to the requirements of cost and performance,hardware modules were selected reasonably.At the same time,according to the requirement of software,the development environments were built,such as ubuntu system,Windows system and Android system.Secondly,the localization technology of mobile robot based on fusion IMU was studied.The problems of angle drift when calculating the attitude and cumulative error when calculating the displacement were improved.After fusing gyro,accelerometer and electronic compass,the attitude angle was calculated by using the complementary filtering.At the same time,the displacement was calculated by using the peaks-troughs algorithm.Thus,the display and location of trajectory in Android system were implemented.Then,the improvement and optimization methods of ORB-SLAM system for mobile robot were studied.The problems of high hardware configuration,time-consuming system and unclear interior structure based on the point cloud were improved.The running speed was improved,when the SLAM system was ported to a low-cost embedded board,the FLANN algorithm was added,RANSAC algorithm was improved and Local BA part was deleted.At the same time,the representation of point cloud was transformed into the representation of raster map.Thus,the interior structure was clearly described and the storage space of map was reduced.Finally,the improvement and optimization of ORB-SLAM system for mobile robot based on fusion IMU were studied.The positioning error of system was improved and the real scale of environment was obtained.The correction of camera poses and graph optimization of G20 based on fusion IMU were implemented.The coordinate of the ORB-SLAM system was corrected by using calman filter which based on fusion IMU.The experiments were verified by the own mobile robot based on Android system.The experimental datas indicated that:By improving and optimizing of the algorithm,the precision SLAM was implemented after porting to embedded board.CPU occupancy rate had been reduced by about 73%.Operation rate was greatly improved.Hardware cost and configuration requirements were greatly reduced.Map stock was reduced by about 98%after the grid map was used,and the actual layout of the environment was reacted more intuitively.The distance error between coordinate of IMU and actual coordinate was controlled within the range of ±0.2m.The location precision greatly improved,when contrasted with the error range of ±0.5m before combining IMU.