Title: Terrain frames classification based on HMC for quadruped robot

Authors: Zhe Li; Yibin Li; Xuewen Rong; Hui Zhang

Addresses: China Water Resources Beifang Investigation, Design and Research Co. Ltd., Tianjin, 300222, China; School of Control Science and Engineering, Shandong University, Jinan, 250061, China ' School of Control Science and Engineering, Shandong University, Jinan, 250061, China ' School of Control Science and Engineering, Shandong University, Jinan, 250061, China ' School of Electrical Engineering and Automation, Qilu University of Technology (Shandong Academy of Sciences), Jinan, Shandong, China

Abstract: As a multi-body nonlinear rigid-flex system, the quadruped robot must maintain the correct perception and classification capabilities for the external environment. This ability is necessary to help quadruped robots make path planning, gait adjustment and attitude control while maintaining complex interactions with the external environment. This paper proposes a terrain classification algorithm based on HMC (HMRF-MAP-CNN) as the basis for robot motion control strategy selection. Different from the classification method based on image features, the terrain-based classification method has higher accuracy and better computational efficiency. In the process of solving the actual terrain classification problem, the algorithm firstly uses HMRF to classify the obtained terrain frames into two categories, flat and rugged, and then use CNN to filter, according to the causes of rugged terrain frames. Through the simulation experiment and comparative analysis, the superiority of HMC terrain frame classification algorithm is confirmed.

Keywords: quadruped robot; terrain classification; Hidden Markov random field; HMRF; Maximum a Posteriori; MAP; convolutional neural network; CNN.

DOI: 10.1504/IJICA.2021.10043061

International Journal of Innovative Computing and Applications, 2021 Vol.12 No.5/6, pp.273 - 280

Received: 04 Apr 2020
Accepted: 21 Apr 2020

Published online: 26 Nov 2021 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article