Accurate perception and representation of rough terrain for a hexapod robot by analysing foot locomotion. (April 2022)

• Record Type: Journal Article
• Authors: Li, Manhong; Wang, Zhe; Zhang, Dong; Jiao, Xin; Wang, Jingtian; Zhang, Minglu
• Is Part Of: Measurement. Volume 193(2022)
• Publication Date: 2022-04
• Subjects: Hexapod robot -- Terrain perception -- Terrain representation -- Foot locomotion; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2022.110904
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: A rough terrain perception and representation method is proposed for a hexapod robot by analysing foot locomotion. A calculation model is constructed to acquire the unordered foot positions by using the vector method of kinematic analysis. A local terrain representation method is proposed by establishing a filter mechanism to integrate foot positions information. A global terrain representation method is presented by using local terrains and the body posture transformation relationship. The proposed method can effectively improve the terrain measurement accuracy without external observation sensors. Abstract: Accurate perception and representation of rough terrain is essential for hexapod robots to perform excellent motion. However, most existing methods currently rely on external observation sensors with low robustness and random error in harsh environments, which generally result in poor terrain measure effect. Inspired by the discrete contact characteristics between feet and terrain during hexapod robot movement, a rough terrain perception and representation method is proposed by analysing foot locomotion. Based on the vector method of kinematic analysis, a calculation model is constructed to dynamically acquire the foot positions during movement. Fully integrating these foot positions information, a local terrain representation method is proposed to precisely depict the robot landing terrain by establishing a feature point filter mechanism. According to the body posture transformation relationship between adjacent motion moments, a global terrain representation method is presented to accurately reconstruct the traversed terrain by cutting and splicing local terrains. Experiment results show the proposed method can accurately perceive and reconstruct the rough terrain travelled by the robot without using external observation sensors. The maximum average perception error in local terrains and the global terrain are reduced to 5.04 mm and 3.28 mm respectively.