SSR: Scaling Surefooted and Symmetric Humanoid Traversal Toward the Open World

This research introduces SSR, an efficient end-to-end framework for egocentric vision-based humanoid traversal. It employs imagined foothold guidance to model forthcoming swing-foot contacts and reduce edge slips, utilizes equivariant latent-space symmetry augmentation for bilateral coordination under high-dimensional visual observations, and incorporates terrain-specific multi-discriminator motion priors to encourage human-like behavior. Extensive experiments demonstrate that SSR achieves safe, stable, and high-quality locomotion on diverse real-world terrains, including stairs with varied structures and extreme challenges such as wide gaps and high platforms, while enabling reliable long-horizon traversal in diverse outdoor environments.