pipette
ENEnglish

Adaptive Traversal Tubes for Posture-Aware Navigation of Bipedal Humanoids

Zijian Ge

PreprintUso en el mundo real

En palabras de los autores

Bipedal humanoids can change their effective footprint by adjusting body yaw, enabling both narrow-passage traversal and socially aware yielding. This paper presents a posture-aware navigation framework centered on Adaptive Traversal Tubes (ATT), Interval Velocity--Heading (IVH) control, and a sidestep-enhanced AMP locomotion prior. ATT represents a behavior-feasible corridor width field along a reference path, unifying geometric clearance and social comfort into a continuous signal. IVH maps this width to a continuous frontal-to-sidestep posture schedule and omnidirectional velocity commands. A velocity-conditioned AMP policy, trained with GVHMR-retargeted frontal/sidestep motions and coupled \((v_x,v_y)\) tracking rewards, is designed to improve transition stability over a generic locomotion controller while remaining interchangeable with the default onboard policy. This preprint presents the full method; a planned evaluation protocol on a Unitree G1 is outlined, with quantitative results to follow. Demo videos of MuJoCo simulation and real-robot deployment are available at: github.com/lborogzj997/Humanoid-Navigation-in-Crowds-and-Narrow-Spaces.

Resultado principalLimitación que admiten los autores

Apareció: miércoles, 21 de octubre. arXiv. Preprint, todavía sin revisión por pares.