Unleashing Humanoid Reaching Potential via Real-world-Ready Skill Space

TL;DR

This paper introduces R2S2, a structured skill space for humanoid robots that enables efficient, zero-shot sim2real transfer for complex reaching tasks by combining primitive skills into a unified latent space.

Contribution

The paper presents a novel skill library and a unified latent space for humanoid reaching, improving transferability and task efficiency over prior methods.

Findings

Zero-shot sim2real transfer achieved in multiple scenarios.

Structured skill space improves control and robustness.

High-level planner effectively utilizes the skill space.

Abstract

Humans possess a large reachable space in the 3D world, enabling interaction with objects at varying heights and distances. However, realizing such large-space reaching on humanoids is a complex whole-body control problem and requires the robot to master diverse skills simultaneously-including base positioning and reorientation, height and body posture adjustments, and end-effector pose control. Learning from scratch often leads to optimization difficulty and poor sim2real transferability. To address this challenge, we propose Real-world-Ready Skill Space (R2S2). Our approach begins with a carefully designed skill library consisting of real-world-ready primitive skills. We ensure optimal performance and robust sim2real transfer through individual skill tuning and sim2real evaluation. These skills are then ensembled into a unified latent space, serving as a structured prior that helps task execution in an efficient and sim2real transferable manner. A high-level planner, trained to sample skills from this space, enables the robot to accomplish real-world goal-reaching tasks. We demonstrate zero-shot sim2real transfer and validate R2S2 in multiple challenging goal-reaching scenarios.