Learning Quiet Walking for a Small Home Robot

2025-02-16

AI Summary

To address excessive footstep noise generated by small quadruped robots in home environments—negatively impacting user experience and social acceptance—this paper proposes a simulation-to-reality (Sim-to-Real) reinforcement learning framework tailored for silent locomotion. The method innovatively optimizes foot-end contact velocity as the primary objective, integrating adaptive PD gain scheduling, foot-force and contact-state feedback, and curriculum learning to achieve controllable trade-offs between acoustic quietness and locomotion robustness. Experimental validation on a physical robot demonstrates that the learned gait significantly reduces acoustic emissions compared to baseline RL approaches: foot-end contact velocity decreases by 32% relative to Sony’s commercial controller, yielding substantial improvements in silence performance while preserving essential motion stability and terrain adaptability.

Technology Category

• Intelligent Robots: Behavior Learning & Control
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Search and Retrieval-Augmented AI: Agentic search
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

As home robotics gains traction, robots are increasingly integrated into households, offering companionship and assistance. Quadruped robots, particularly those resembling dogs, have emerged as popular alternatives for traditional pets. However, user feedback highlights concerns about the noise these robots generate during walking at home, particularly the loud footstep sound. To address this issue, we propose a sim-to-real based reinforcement learning (RL) approach to minimize the foot contact velocity highly related to the footstep sound. Our framework incorporates three key elements: learning varying PD gains to actively dampen and stiffen each joint, utilizing foot contact sensors, and employing curriculum learning to gradually enforce penalties on foot contact velocity. Experiments demonstrate that our learned policy achieves superior quietness compared to a RL baseline and the carefully handcrafted Sony commercial controllers. Furthermore, the trade-off between robustness and quietness is shown. This research contributes to developing quieter and more user-friendly robotic companions in home environments.

Problem

• Minimize footstep noise in home robots
• Implement reinforcement learning for quiet walking
• Balance robot robustness and quietness

Innovation

• Sim-to-real reinforcement learning
• Varying PD gains
• Foot contact sensors