Research On Positioning And Navigation Technology Of Mobile Robot Based On Embedded Platform

Mobile robots are widely used in medical services,industrial production and homes services.The various basic functions of mobile robot require that it can be controlled at long range,and at the same time the movement of mobile robot in the scene requires the ability of perceiving the current environment accurately and autonomous navigation.So it is of great significant to research the mobile robot function of positioning and navigation under remote control.This paper designs the robot remote distributed control platform based on embedded board,which PC communicate with embedded board through WIFI,and the embedded receives and sends data with mobile robot and sensors through serial port and USB.The PC's control of robot's bottom motion system and the reception of the sensor data are realized by the Robot Operating System.In order to realize the recognition of mobile robot in indoor environment and further improve the scene positioning accuracy on the basis of traditional scene classifier,this paper proposes a positioning method of mobile robot in indoor environment based on Q-learning,which solves the low positioning accuracy problem when the detection range is small.For the problem of fitting Q-learning action value function,we propose a neural network algorithm that combining Extreme learning machine and back propagation gradient descent,which reduces the computational cost of neural network.For the problem that the classification accuracy of the scene profile image is low,we use the feature extraction method of ring projection to deal with the scene profile image,and then improve the classification accuracy.The mobile robot can learn the robot orientation angle with the highest scene recognition rate dynamically based on Q-learning through the algorithm,so that the robot can acquire more reliable sensor information and then fuse the corresponding recognition results,thus improving the accuracy of scene positioning.The algorithm is applied to the experiment of mobile robot scene positioning,and the results show that the algorithm can effectively improve the accuracy of scene positioning.In order to realize autonomous movement function of mobile robot,we propose a mobile robot navigation method based on fuzzy control and Q-learning,which realizes the path planning function of mobile robot.The mobile robot can set the target autonomously and approach the target under the fuzzy control rule.Through training the Q-learning neural network in the scene,the mobile robot can avoid the obstacle in the process of approaching the target.Set up the simulation scene and get the obstacle avoidance training in there,and then get the navigation experiment.Experimental results show that mobile robots have the ability of autonomous navigation.