Adaptive Foothold Selection for a Hexapod （自适应选择立足昆虫）.pdf

Adaptive Foothold Selection for a Hexapod Robot Walking on Rough Terrain D. Belter ∗ ∗ Institute of Control and Information Engineering, Poznan University of Technology, Poznan, Poland (e-mail: Dominik.Belter@put.poznan.pl). Abstract: This paper introduces an adaptive foothold selection algorithm for a legged robot walking on a rough terrain. The proposed algorithm searches for the best footholds to prevent slippages. It uses a known grid map of the surrounding terrain and a polynomial-based approximation method to create a decision surface. The robot learns from experiments, therefore no a priori expert’s knowledge is required. The results of simulations show that the method is general enough to work properly on various types of terrain. These results have been also verified on the real walking robot ’Ragno’. Keywords: walking robot, motion planning, learning, approximation 1. INTRODUCTION 1.1 Problem statement Walking robots are in the area of high interest because of their potential abilities to deal with a rough terrain. They can be successfully applied in search and rescue missions. However, walking on an unstructured terrain requires an appropriate control algorithm. In such an algorithm there are three main problems to deal with: foothold selection, static and dynamic stability, and path planning. The foothold selection method is crucial, because whenever the feet are placed improperly on the ground the probability of slippage and the risk of fall are much higher. On the other hand, a