Mobile Robot Traversability Prediction Based on Differential Evolution Support Vector Machine

摘要

A traversability prediction method for the mobile robot is presented.The traversability is described with the relative vibration.The color and texture features,which are one component of training sample,are extracted from the images of typical terrain.And the relative vibration intensity,which is the other component of training sample,is measured when the mobile robot travels across the typical terrain.The fitting method between training samples and relative vibration intensity based on differential evolution support vector machine is developed,of which the model parameters are optimized with differential evolution algorithm.When the mobile robot works,the image of the front terrain is linearly divided into sub-regions for prediction.The color and texture features of the sub-regions are extracted and the traversability values of the sub-regions are predicted based on the trained differential evolution support vector machine.For the smoothness of movement,the optimal path method is given based on traversability prediction with distance coefficient.The experiment demonstrates that the method can give an effective traversability prediction of the complex terrain for the mobile robot.