Research And Implementation Of Laser And Vision Fusion SLAM Algorithm Applied To Robots

With the popularity of the autonomous driving and robot industry,simultaneous localization and mapping(SLAM)of mobile robots has entered the stage of rapid development.RTAB-Map uses an RGB-D camera and a 2D laser to realize a laser and vision fusion SLAM system with low cost and high accuracy.However,when the RTAB-Map SLAM system is in constant motion,it will cause the camera and inertial measurement unit(IMU)initialization failure in the front end of the system and the cumulative error of the vision-inertial tightly coupled odometery to increase.Therefore,this article optimizes the system initialization and odometery parts in the front end of the SLAM system.The main research work of this paper is as follows:1.Aiming at the problem that the IMU acceleration observation is zero in the process of uniform motion,resulting in the failure of the initialization of the system,an optimization algorithm based on the camera and IMU initialization assisted by encoder observation is proposed.The algorithm mainly uses wheel encoder translation observation to assist in estimating the initial velocity and gravity vector of the IMU.It solves the problem that the joint initialization of the camera and the IMU fails due to insufficient IMU excitation in the process of uniform motion,which improves the accuracy and robustness of the joint initialization of the camera and the IMU.Moreover,on the basis of camera and IMU initialization,the camera and wheel encoder initialization external parameter calibration method is integrated.2.Aiming at the problem that the accuracy of the visual-inertial tightly coupled odometery decreases when the indoor mobile robot moves at a constant speed,a vision-inertial tightly coupled odometery based on wheeled encoder is proposed.The observation of wheel encoder extends the IMU pre-integration to the encoder-IMU pre-integration,which is tightly coupled with visual observation to jointly constrain the posture of the mobile robot.In addition,the external parameter online calibration between the camera,IMU and wheel encoder is realized.The system optimization model of the fusion odometery,the residual of the objective function,and its Jacobian matrix are derived.It is verified that the fusion odometery scheme has higher accuracy in posture tracking during constant speed movement,and is more suitable for indoor mobile robots.3.The implementation of the software and hardware platform of the laser and vision fusion SLAM system,and the relevant tests and experiments are carried out in the real environment.It is verified that the laser and vision fusion SLAM system is more complete than the laser SLAM system.This paper verifies that the fusion odometery proposed in this paper can improve the mapping accuracy of laser and vision fusion SLAM systems.