A supervisory control approach for Digit navigation on inclined terrain
Date
2024
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
Humanoid robots are engineered to replicate human capabilities, including navigating inclined terrains with stable locomotion. Humans exhibit remarkable adaptive capability to safely traverse inclined terrains without sight, adjusting their gait dynamically based on proprioceptive feedback and tactile sensations. Therefore, to enable bipedal robot Digit to navigate on inclined terrains without sight, we have implemented a high level planner called Supervisory Control System. ☐ In this thesis, we have proposed the development and implementation of the Supervisory Control System capable of operating in the presence of potentially large parametric uncertainties, specifically addressing the inclination angle of the terrain. Supervisory Control is designed as a combination of an online estimation unit and a controller library, where a suitable controller is selected based on the current estimate of ground inclination and the system behavior is adapted accordingly. In this control design, optimal controller is selected by the means of logic-based switching among the family of controllers and is placed in the feedback loop based on the current estimation of the inclination angle. We illustrate the effectiveness of our method by applying it on Digit walking on inclined terrain. Our results show that by switching to the optimal controller that is best according to the current estimate of uncertainty maintains robustness and stability across varying inclinations.
Description
Keywords
Humanoid robots, Locomotion, Supervisory Control System, Logic-based switching, Tactile sensations