TAMOLS: Terrain-Aware Motion Optimization for Legged Systems

Terrain geometry is, in general, nonsmooth, nonlinear, nonconvex, and, if perceived through a robot-centric visual unit, appears partially occluded and noisy. This article presents the complete control pipeline capable of handling aforementioned problems real-time. We formulate trajectory optimization problem that jointly optimizes over base pose footholds, subject to height map. To avoid converging into undesirable local optima, we deploy graduated technique. embed compact, contact-force free stability criterion is compatible with nonflat ground formulation. Direct collocation used as transcription method, resulting nonlinear can be solved online less than ten milliseconds. increase robustness presence external disturbances, close tracking loop momentum observer. Our experiments demonstrate stair climbing, walking on stepping stones, gaps, utilizing various dynamic gaits.