Model Predictive Control for Legged Robots

This PhD thesis investigates the role of optimal planning in autonomous legged locomotion. It addresses dynamic modeling choices, trajectory optimization limitations, and proposes a robust NMPC framework validated on physical quadruped platforms.

Key Research Contributions:

• Dynamic Model Validation: Closed-loop validation of Single Rigid Body Dynamics (SRBD) models.
• LTV Trajectory Optimization: Introduction of a Linear Time-Varying based trajectory optimization scheme for legged locomotion.
• Real-Time NMPC: Development of an NMPC with a novel mobility-based criterion, enhancing adaptation to complex terrain features.
• Experimental Validation: Validated extensively on the HyQ quadruped robot for omni-directional walking and rough terrain navigation.