Feasible Wrench Based Locomotion Strategy for Legged Robots

The field of legged robotics has seen impressive developments in the past decade. Significant progress has been made in developing various approaches that enable robots to achieve highly dynamic and stable motions. Legged robots have been used successfully in the industry for inspection applications. However, the potential to use legged robots in interactive applications where the robot can be pushed to its limits is still considerable. In such applications, feasible locomotion under the stringent constraints of terrain interaction and the robot's actuation limits is needed. The objective of this thesis is to present a novel locomotion strategy based on base wrench requirements and feasibility guarantees, which aims to enhance the ability of legged robots to execute complex interactive maneuvers by maximizing the achievable wrenches within the robot's capabilities. The first part of this thesis analyzes different feasibility measures. It develops the feasible region and the feasibility margin to incorporate our wrench requirements, stability and contact constraints, and actuation limits. We further use machine-learning models to approximate the feasibility margin and provide algorithmic differentiation possibility to compute its gradients. In the second part of the thesis, we leverage the feasibility criteria in developing motion planning strategies. We design trajectory planning methods for the base of the robot as well as for foothold selection. We show that using such simplified criteria can be efficiently used in both improved heuristic and gradient-based optimization approaches. Simulation and experimental results, demonstrated using the Aliengo, HyQ, and HyQReal robots, validate the effectiveness of the feasibility margin based strategies in achieving robust locomotion under significant external disturbances and on challenging terrains. Comparisons with traditional euristic and optimization strategies highlight the importance of considering motion and actuation feasibility in ensuring more stable and feasible locomotion.