The Design And Implementation Of The Usher Robot Navigation System Using Multi-sensor Information Fusion

In recent years,with the continuous development of sensor technology and artificial intelligence,robot technology has been rapidly innovated and widely applied.The traditional service robot cannot meet the expectation of intelligent life,so it has become a requirement to develop a service indoor robot with autonomous navigation function.The research shows that excellent positioning technology and path planning technology is the key technology in the field of robot navigation.In order to achieve good accuracy and stability of the navigation function,this paper combined with open source robot operating System(Robot Operation System,ROS),through the extended Kalman filtering algorithm fusion encoder,inertial measurement unit and monocular camera positioning information,completely designed and built the service robot navigation system.The main research contents of this paper are as follows:1.The research of monocular camera vision odometry.The general framework of monocular camera vision odometer was analyzed in detail,including feature point extraction and matching,camera motion estimation and local optimization.This chapter compared the open source Mono SLAM and ORBSLAM2 algorithm based on the feature point method,feature point extraction,SLAM mapping results and positioning accuracy difference,through the mean square error evaluation of the two algorithms in the difference in positioning accuracy.2.Construction of odometry integrating multi-sensor information.In the construction of multi-sensor information fusion algorithm model,the estimation of the encoder was taken as the system model,and the estimation of the monocular camera and the inertial measurement unit was taken as the observation value to update the position information and attitude information respectively.Using the existing hardware on the robot,the position fusion experiment and attitude fusion experiment are carried out.Experimental results showed that the fusion pose information is more accurate than that of a single sensor.3.Research on path planning algorithm based on the grid map.This chapter introduced the basic principle and search process of the path planning algorithm based on the grid map.Aiming at the problem that the current A~* algorithm does not take into account the robot width information and the uneven trajectory planning,an improved strategy was proposed.In the estimation function,the safety factor of obstacle nodes is considered,and the children nodes which are in the same direction as the original path are generated first.In the generated path planning sequence,the connecting lines of front and rear nodes are removed without passing through the intermediate nodes of obstacles,so as to smooth the trajectory.Simulation experiments was designed to verify the difference between the original algorithm and the improved algorithm in trajectory planning.4.Designed and built the navigation system of the service robot.The hardware design and software design of the sevice robot are discussed in detail.Combined with the ROS of the robot operating system,the improved A~* algorithm is integrated into the Move Base navigation framework by the method of Plugin.The differences in path planning between the standard A~* algorithm and the improved A~* algorithm were compared in the indoor scene and corridor scene,and twelve times navigation experiments were conducted in the corridor scene to verify the stability and accuracy of the navigation system of the service robot.In the navigation experiment,the average error is stable around 10 cm,and the service robot has good navigation accuracy and practical application prospect.