A Modular Residual Learning Framework to Enhance Model-Based Approach for Robust Locomotion

2025-07-24

AI Summary

Model mismatch severely degrades controller robustness under parametric uncertainties and environmental disturbances. Method: This paper proposes a modular residual learning framework that synergistically integrates model-based components—dynamic modeling, model-based gait planning, and model predictive control—with data-driven learning. Crucially, lightweight residual networks are embedded at key stages of gait planning and dynamics modeling, with learning strategies customized to each component’s characteristics to improve training efficiency and generalization. These residual modules are end-to-end optimized via deep reinforcement learning to achieve high-fidelity compensation in simulation. Contribution/Results: The resulting hybrid controller significantly enhances the nominal controller’s robustness against parameter variations and external disturbances. Experiments on a real quadrupedal robot demonstrate stable locomotion and precise velocity tracking. Notably, the method maintains balance and reliably executes motion commands even under strong unmodeled uncertainties—such as unknown terrain and actuator delays—beyond the scope of the simulation model.

Technology Category

• Intelligent Robots: Behavior Learning & Control
• Machine Learning: Adversarial Learning & Robustness
• Planning, Routing, and Scheduling: Learning for Planning and Scheduling

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

This paper presents a novel approach that combines the advantages of both model-based and learning-based frameworks to achieve robust locomotion. The residual modules are integrated with each corresponding part of the model-based framework, a footstep planner and dynamic model designed using heuristics, to complement performance degradation caused by a model mismatch. By utilizing a modular structure and selecting the appropriate learning-based method for each residual module, our framework demonstrates improved control performance in environments with high uncertainty, while also achieving higher learning efficiency compared to baseline methods. Moreover, we observed that our proposed methodology not only enhances control performance but also provides additional benefits, such as making nominal controllers more robust to parameter tuning. To investigate the feasibility of our framework, we demonstrated residual modules combined with model predictive control in a real quadrupedal robot. Despite uncertainties beyond the simulation, the robot successfully maintains balance and tracks the commanded velocity.

Problem

• Combines model-based and learning-based frameworks for robust locomotion
• Addresses performance degradation from model mismatch via residual modules
• Enhances control in uncertain environments with modular learning methods

Innovation

• Combines model-based and learning-based frameworks
• Modular residual learning enhances robustness
• Integrates model predictive control in robots