Back to Projects

Model Predictive Control for Legged Robots

PhD Thesis Legged Locomotion MPC Robotics

A comprehensive thesis on optimal planning for legged locomotion, featuring a real-time NMPC with mobility-based cost function for rough terrain traversal.

Thesis Overview

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.

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