Adaptation of a Decentralized Controller to Curve Walking in a Hexapod Robot

Simmering, J; Hermes, L; Schneider, A; Schilling, M

Abstract

Curve walking is an important skill for multi-legged robot locomotion as it increases a robots' maneuverability. We investigate whether a decentralized system can explain even complex walking behaviors like curve walking. Based on an analysis of the curve walking capabilities of the decentralized control architecture Walknet, we propose a couple of simple but effective modifications with a main focus on the coordination between leg controllers: controlling the step length, shifting the AEP (transition point from swing to stance), and decreasing the step length for the legs on the inside of the curve. In simulation, the modified architecture demonstrated a significant improvement in the stability of the curve walking performance for tight curves and even allows a smooth transition to extremely tight curves and turning on the spot. Furthermore, the system is tested on a real robot and showed good qualitative results and robust curve walking behavior.