Research On Key Technologies Of Autonomous Navigation System For Service Robot

With the development of society,the service robot plays a more and more important role in people's daily life.Autonomous navigation technology is the basic guarantee for robot to achieve human-machine interaction and complete tasks.This paper mainly studies the key technologies of autonomous navigation system for service robots.The overall framework of service robot navigation system is designed,including external perception unit,posture acquisition unit,central processing unit,bottom control unit and motor drive unit.The Kinect sensor is used to simulate the laser sensor to obtain the laser data,and a multi-layer laser scanning method based on the height of the robot is designed.In addition,a laser slam system fusing the visual information is studied.Aiming at the large error of traditional encoder odometry when obtaining data,visual odometry is designed to provide a pose information for slam.The ICP algorithm and the polar geometric algorithm commonly used to construct visual odometer are studied.Aiming at the situation that the depth data error of the Kinect sensor affects the accuracy of visual odometry,a depth error model is constructed and added to the visual odometer to improve the accuracy of pose calculation.Because the laser sensor can only capture a plane geometric information,is not easy to find in the loop detection based on laser SLAM backend optimization in vision sensor.While the visual sensor can achieve rich information,it is effective by adding the visual loop closure detection into the laser slam system to improve the success rate of loop closure detection.Then the autonomous navigation technology of service robot based on slam map is studied.The location of the robot based on the known map is realized by using the Monte Carlo location algorithm,and the number of particles is adjusted adaptively by the KLD sampling algorithm.Aiming at the situation when dynamic obstacle blocks the path and robot can not make a path planning,joining a recovery behavior to the navigation system.Experiments were conducted and showed that when dynamic obstacles block the path,the robot can also make a path planning.Finally,Through the gazebo to build the experimental environment,Experiment showed that the SLAM algorithm incorporating the visual loop closure detection caneffectively find the loop closure point and control the accumulation of errors.The experiment of path planning based on SLAM map showed that the robot can plan the path effectively and follow the planning path.