Design Of Wheeled Robot Adaptive Odometer Based On Terrain Classification

Wheeled robots are widely used in various industries due to their simple structure and stable performance.However,in the current work demand,wheeled robots are often in complex terrain environments,where wheel sliding and irregular wheel surfaces can lead to odometer errors.This error can only be suppressed through probability methods and cannot be completely eliminated.Therefore,it is of great significance to improve the odometer accuracy of wheeled robots in complex terrain.The main work of this article is as follows:(1)In order to solve the problem that the odometer of the current robot is more easily divergent in complex terrain,and the terrain recognition method based on the depth network of the robot has a long network training time and large training data,a terrain recognition method based on the depth residual network and transfer learning is proposed.This method first establishes six typical flat terrain image datasets;Secondly,a new deep residual network is constructed based on the Resnet network,and the Imagenet dataset is used to pre train the constructed deep residual network.The feature transfer method is used to achieve large-scale parameter transfer of the pre trained network;Finally,six terrain image datasets were used to train the fully connected layer of the deep residual network.The terrain recognition method has an average recognition accuracy of 99.3% for six types of terrain.(2)In order to solve the problem of wheeled robot sliding in complex terrain,firstly,based on the traditional wheeled robot kinematics model,the kinematics model when the wheel slips is established,and the factor affecting the odometer accuracy of wheeled robot in complex terrain environment is determined as the resistance coefficient under different terrain conditions;Then,the resistance coefficient under different terrains is determined through experiments using wheeled robots.Considering the impact of different terrains on the accuracy of robot odometer estimation,an adaptive odometer system based on terrain classification was designed.This system can independently modify the resistance coefficient of wheeled robots based on the current terrain type,achieving the adaptability of odometers.(3)The wheeled robot is used to simulate the complex terrain indoors and the actual complex terrain outdoors.The results show that the mean odometer displacement error and course angle error of the wheeled robot based on the adaptive odometer system are 0.151 m and 0.024 rad in indoor artificial simulation of complex terrain.The mean displacement error of odometer is 0.403 m and the mean heading angle error is 0.036 rad under the actual outdoor complex terrain.