RPL: Learning Robust Humanoid Perceptive Locomotion on Challenging Terrains

2026-02-03

AI Summary

This work addresses the challenge of achieving robust omnidirectional locomotion for humanoid robots on complex terrains, particularly under payload conditions. The authors propose RPL, a two-stage training framework: first, terrain-specific expert policies are trained using privileged height maps; then, these policies are distilled into a vision-based Transformer policy that relies solely on multi-view depth cameras. Key innovations include velocity-command-driven depth feature scaling and a randomized lateral masking mechanism to enhance adaptability to asymmetric observations and unknown terrain widths. An efficient multi-depth simulation system is developed, integrating ray-casting acceleration, parallel rendering of static and dynamic meshes, and realistic modeling of sensor noise and latency. Real-world experiments demonstrate that the robot robustly traverses challenging terrains—including 20° slopes, variable-step stairs, and 25 cm stepping stones spanning 60 cm gaps—while carrying a 2 kg payload.

Technology Category

• Intelligent Robots: Motion and Path Planning
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Humanoid perceptive locomotion has made significant progress and shows great promise, yet achieving robust multi-directional locomotion on complex terrains remains underexplored. To tackle this challenge, we propose RPL, a two-stage training framework that enables multi-directional locomotion on challenging terrains, and remains robust with payloads. RPL first trains terrain-specific expert policies with privileged height map observations to master decoupled locomotion and manipulation skills across different terrains, and then distills them into a transformer policy that leverages multiple depth cameras to cover a wide range of views. During distillation, we introduce two techniques to robustify multi-directional locomotion, depth feature scaling based on velocity commands and random side masking, which are critical for asymmetric depth observations and unseen widths of terrains. For scalable depth distillation, we develop an efficient multi-depth system that ray-casts against both dynamic robot meshes and static terrain meshes in massively parallel environments, achieving a 5-times speedup over the depth rendering pipelines in existing simulators while modeling realistic sensor latency, noise, and dropout. Extensive real-world experiments demonstrate robust multi-directional locomotion with payloads (2kg) across challenging terrains, including 20{\deg} slopes, staircases with different step lengths (22 cm, 25 cm, 30 cm), and 25 cm by 25 cm stepping stones separated by 60 cm gaps.

Problem

• humanoid locomotion
• challenging terrains
• multi-directional locomotion
• robust perception
• payload handling

Innovation

• perceptive locomotion
• two-stage training
• depth distillation
• transformer policy
• multi-depth simulation