Logo image
Leveraging Stability Analysis for Robust Humanoid Motion Planning and Teleoperation
 

Leveraging Stability Analysis for Robust Humanoid Motion Planning and Teleoperation

Stephen McCrory
University of West Florida Libraries
Doctor of Philosophy (PHD), University of West Florida
2026
1
Humanoid robots are designed with the aim of navigating difficult terrains at a capability that matches or even exceeds humans. A central problem of motion planning for humanoid robots in complex environments is high dimensionality. In deciding contacts, the robot must consider numerous sensor inputs along with an internal model of its interaction capabilities. Broadly, in this thesis we develop algorithms that can address this by accepting rich map representations and efficiently reasoning about contact stability. We begin this thesis by introducing a novel approach to modeling bipedal traversability, which encodes the bipedal gait directly into a graph structure. This enables sample-based planning at the navigation level through A* graph search. Next, we apply previous work in multi-contact centroidal stability in two teleoperation contexts. The first is a command-based teleoperation scheme that provides the operator with visual feedback of the robot’s stability, allowing the operator to optimize contact and posture. The second is a motion-based teleoperation scheme that provides automatic adjustment of the operator’s posture and contacts to improve stability. Finally, we present a hand contact planner that enables the robot to reactively brace while both standing and walking, based on a learned model of centroidal stability. We demonstrate that these approaches are deployable to real-world hardware through a number of humanoid platforms. This includes demonstrations such as walking across rough terrain (cinder blocks, stairs, ramp), transitions between postures of mixed feet, hand and knee contacts as well as bracing while grasping and reactively bracing while walking. Through these demonstrations, along with corresponding simulations, we aim to highlight the robot’s robustness and the speed of the backing algorithms as the main success indicators.

(1)

pdf
Leveraging Stability Analysis for Robust Humanoid Motion Planning and Teleoperation25.01 MB
Preprint Dissertation pdf Open Access
Logo image