LEGO: Latent-Space Exploration for Geometry-Aware Optimization of Humanoid Kinematic Design
JiHwan Yoon, Taemoon Jeong, Jeongeun Park, Chanwoo Kim, Jaewoon Kwon, Yonghyeon Lee, Kyungjae Lee, Sungjoon Choi
AI summary
Problem
Robot kinematic design traditionally relies on human intuition due to the vast, unstructured design space and the difficulty of defining task-specific loss functions.
Approach
The method learns a compact, isometric latent space from a dataset of real robot designs using screw-theory representation, then optimizes kinematic structures in this space using gradient-free search guided by motion retargeting and Procrustes analysis against human motion.
Key results
- Screw-theory representation enables scalable encoding of joint axes and positions.
- Isometric autoencoder learns a geometry-preserving latent space from 30 real robot designs.
- Gradient-free latent optimization discovers parsimonious upper-body structures tailored to specific human motions.
- Framework successfully bridges human motion data and robot kinematics without manual loss engineering.
Why it matters
It enables automated, data-driven robot morphology design that reduces manual engineering effort and improves task-specific performance.
Abstract
Designing robot morphologies and kinematics has traditionally relied on human intuition, with little systematic foundation. Motion–design co-optimization offers a promising path toward automation, but two major challenges remain: (i) the vast, unstructured design space and (ii) the difficulty of constructing task-specific loss functions. We propose a new paradigm that minimizes human involvement by (i) learning the design search space from existing mechanical designs, rather than hand-crafting it, and (ii) defining the loss directly from human motion data via motion retargeting and Procrustes analysis. Using screw-theory-based joint axis representation and isometric manifold learning, we construct a compact, geometry-preserving latent space of humanoid upper body designs in which optimization is tractable. We then solve design optimization in this latent space using gradient-free optimization. Our approach establishes a principled framework for data-driven robot design and demonstrates that leveraging existing designs and human motion can effectively guide the automated discovery of novel robot design. Project page: https://jihwan-yoon-page.github.io/legopt/