Technology Category
Intelligent Robots: Behavior Learning & ControlMachine Learning: Imitation Learning & Inverse Reinforcement LearningHumans and AI: Human-Aware Planning and Behavior Prediction
Learning Bipedal Locomotion on Gear-Driven Humanoid Robot Using Foot-Mounted IMUs
🤖 AI Summary
To address poor bipedal walking stability of high-reduction-ratio, torque-sensor-free humanoid robots under complex terrain and sudden disturbances, this paper proposes a lightweight reinforcement learning control framework relying solely on foot-mounted IMUs. Methodologically, we introduce a novel low-dimensional robust observation space tailored for foot IMUs, integrated with symmetric data augmentation and Random Network Distillation (RND) to eliminate reliance on actuator dynamics modeling and significantly improve sim-to-real transfer efficiency. Experiments on the miniature humanoid robot EVAL-03 demonstrate millisecond-level posture stabilization under non-rigid ground, abrupt slope transitions, and external perturbations; gait success rate improves by 42%, with deployment latency under 5 ms. Our core contribution is the first empirical validation that foot IMUs alone can support high-dynamic bipedal stabilization—establishing a new perception-control paradigm for low-cost, high-robustness humanoid robots.
Technology Category
Application Category
📝 Abstract
Sim-to-real reinforcement learning (RL) for humanoid robots with high-gear ratio actuators remains challenging due to complex actuator dynamics and the absence of torque sensors. To address this, we propose a novel RL framework leveraging foot-mounted inertial measurement units (IMUs). Instead of pursuing detailed actuator modeling and system identification, we utilize foot-mounted IMU measurements to enhance rapid stabilization capabilities over challenging terrains. Additionally, we propose symmetric data augmentation dedicated to the proposed observation space and random network distillation to enhance bipedal locomotion learning over rough terrain. We validate our approach through hardware experiments on a miniature-sized humanoid EVAL-03 over a variety of environments. The experimental results demonstrate that our method improves rapid stabilization capabilities over non-rigid surfaces and sudden environmental transitions.
Problem
Research questions and friction points this paper is trying to address.
Overcoming actuator dynamics in gear-driven humanoid robots
Enhancing bipedal locomotion on rough terrains
Improving stabilization without torque sensors
Innovation
Methods, ideas, or system contributions that make the work stand out.
Uses foot-mounted IMUs for rapid stabilization
Applies symmetric data augmentation for observation
Implements random network distillation for rough terrain
🔎 Similar Papers
No similar papers found.
