Skip to content
Conference

Reinforcement Learning-based Impact Forces-Optimized Gait Control for Humanoid Locomotion: Toward Smooth and Quiet Walking

Jul 2026 · 2026 23rd International Conference on Ubiquitous Robots (UR) · pp. 145-149 · 0 citations · 35 references

Abstract

This paper proposes a reinforcement learning-based Impact Forces-Optimized Gait Control framework for achieving smooth and quiet humanoid locomotion. The proposed method integrates a quintic polynomial swing-foot trajectory designed to ensure zero vertical velocity and acceleration at foot-ground contact with a joint-jerk minimization reward. Impact forces are provided as privileged observations during training to improve robustness under an asymmetric information architecture. The effectiveness of the proposed method was verified through evaluations conducted in Isaac Lab and MuJoCo (Multi-Joint dynamics with Contact), demonstrating reduced impact forces across various locomotion modes. Additionally, velocity tracking evaluations confirm that the proposed method does not significantly degrade locomotion performance. Furthermore, its performance was validated on the G1 humanoid robot, where real-world walking tests confirmed a reduction in walking-induced acoustic noise.

View source

Similar papers

Jul 2026

Impact-aware compliant foot placement for quiet humanoid locomotion via reinforcement learning

: Recent advances in legged robots have substantially improved their locomotion capabilities over outdoor and complex terrains. However, quiet locomotion for humanoid robots in noise-sensitive indoor environments remains underexplored, despite its growing importance in human-centered applications. While encouraging progress has been made in quadrupedal robots, transferring the quiet locomotion ability to humanoid robots remains nontrivial due to their fundamentally different foot-ground contact patterns. This paper proposes a control method for reducing foot–ground contact noise during humanoid walking, achieving compliant contact and continuous regulation of locomotion noise by establishing a foot corner contact model along with a virtual compliance parameter. The experimental results show that the average sound pressure level is reduced by 4.88 dB.

Kaifeng Jian, Shaojie Zhang, Dong Zhu et al. · 0 citations
Open access Aug 2026

Reinforcement Learning-Based Interactive Control of an Omnidirectional Mobile Lower Limb Rehabilitation Robot

The findings indicate that integrating reinforcement learning with Sigmoid parameter adaptation provides a systematic and effective solution for adaptive compliance regulation in mobile exoskeleton systems, enhancing adaptability, safety, and functional relevance for stroke patients undergoing lower limb rehabilitation training.

Suyang Yu, Yangqing Yu, Changlong Ye · 0 citations
Preprint Aug 2026

Learning Fault-Tolerant Locomotion with Adaptive Gait Timing

Hardware failures require legged robots to rapidly reorganize coordination and gait timing to maintain stability and mobility. This is particularly challenging for larger quadrupeds, where increased mass and tighter actuation limits reduce the feasibility of aggressive, high-frequency compensation strategies often observed on smaller platforms. In this work, we propose a deep reinforcement learning approach for fault-tolerant locomotion under actuator power loss. The method employs an asymmetric actor-critic architecture in which the critic has access to privileged information during training, while the actor learns to reconstruct a corresponding latent representation from proprioceptive observations. We introduce a latent-alignment loss that encourages consistency between actor and critic representations. Additionally, we augment the action space with a learnable gait frequency parameter, enabling adaptive gait timing in response to terrain variations and actuator degradation without predefined faulty-leg strategies. The approach is validated in high-fidelity simulation on uneven terrain and real-world experiments on flat ground using a 68 kg quadruped robot.

Giovanbattista Gravina, Luca Rossini, Carlo Rizzardo et al. · 0 citations
Conference Jul 2026

Learning dynamically stable Ultra-Slow Gait in Quadruped Robots

This research proposes a novel control architecture for dynamically stable ultra-slow quadruped locomotion on unstructured and slippery terrains. The approach is based on Imitation Learning (IL), where two neural controllers are trained to imitate an optimal Quadratic Programming (QP) controller derived from a Linear Time Invariant (LTI) Variation-Based Linearized (VBL) model of the robot. Stability and safety are incorporated through Lyapunov-inspired and Linear Matrix Inequality (LMI) constraints integrated into the training process. A switching strategy between leg pairs enables continuous balance during slow gait execution. Preliminary simulation results demonstrate accurate center-of-mass tracking and stable behavior across gait phases, supporting the feasibility of the proposed method for real-world deployment.

Alex Li Noce, P. Manoonpong, L. Patané et al. · 0 citations
Conference Jul 2026

Underactuated Virtual Gravity Control: Harnessing Passive Dynamics for Optimally Efficient Bipedal Locomotion

This paper explores the Underactuated Virtual Gravity (UVG) controller, a model-based control approach designed to achieve energy-efficient bipedal walking. The UVG controller minimizes actuator effort during level-ground walking by capitalizing on the inherent dynamics that facilitate stable passive gaits on downward slopes. By effectively leveraging torso dynamics to support the application of the UVG, the method turns the innate underactuation of bipedal systems into an advantage. Extensive benchmarking against state-of-the-art methods such as the Trajectory Optimization for trajectory planning combined with Non-linear Model Predictive Control for tracking shows that the UVG achieves superior energy efficiency within its effective range while constituting a closed-form controller that requires fewer computational resources than numerical methods. The results highlight the UVG and related dynamics-based approaches as a compelling option for energy-conscious, task-focused robot designs.

Aikaterini Smyrli, E. Papadopoulos · 0 citations