ALLSTEPS: Curriculum-driven Learning of Stepping Stone Skills

2020/05/09 by Xie, Zhaoming, Ling, Hung Yu, Kim, Nam Hee +1 · 9 citations
#FOS: Computer and information sciences #Graphics (cs.GR) #Machine Learning (cs.LG) #Robotics (cs.RO)

paper · doi:10.48550/arxiv.2005.04323

Abstract

Humans are highly adept at walking in environments with foot placement constraints, including stepping-stone scenarios where the footstep locations are fully constrained. Finding good solutions to stepping-stone locomotion is a longstanding and fundamental challenge for animation and robotics. We present fully learned solutions to this difficult problem using reinforcement learning. We demonstrate the importance of a curriculum for efficient learning and evaluate four possible curriculum choices compared to a non-curriculum baseline. Results are presented for a simulated human character, a realistic bipedal robot simulation and a monster character, in each case producing robust, plausible motions for challenging stepping stone sequences and terrains.

Cited by

• Towards Senior-Robot Interaction: Reactive Robot Dog Gestures
• Learning Terrain Aware Bipedal Locomotion via Reduced Dimensional Perceptual Representations
• FreeMusco: Motion-Free Learning of Latent Control for Morphology-Adaptive Locomotion in Musculoskeletal Characters
• Sim-to-Real Transfer in Deep Reinforcement Learning for Bipedal Locomotion
• Learning to Ball: Composing Policies for Long-Horizon Basketball Moves
• HuMam: Humanoid Motion Control via End-to-End Deep Reinforcement Learning with Mamba
• Traversing Narrow Paths: A Two-Stage Reinforcement Learning Framework for Robust and Safe Humanoid Walking
• RESCUE: Crowd Evacuation Simulation via Controlling SDM-United Characters
• Hierarchical Reinforcement Learning and Value Optimization for Challenging Quadruped Locomotion

Related

• Learning Dynamic Bipedal Walking Across Stepping Stones
• Sim-to-Real Learning of Footstep-Constrained Bipedal Dynamic Walking
• Learning Bipedal Walking On Planned Footsteps For Humanoid Robots
• Towards Quadrupedal Jumping and Walking for Dynamic Locomotion using Reinforcement Learning
• Adaptive Step Duration for Accurate Foot Placement: Achieving Robust Bipedal Locomotion on Terrains with Restricted Footholds