Perceptive Variable-Timing Footstep Planning for Humanoid Locomotion on Disconnected Footholds

TL;DR

This paper introduces a perceptive, real-time footstep planning method for humanoid robots that navigates disconnected footholds by integrating depth perception, convex region extraction, and safety bounds into a mixed-integer predictive control framework.

Contribution

It presents a novel onboard planning approach combining perception, convex region extraction, and mixed-integer optimization for safe, adaptive humanoid locomotion on complex terrains.

Findings

Planner achieves millisecond-level solve times.

Successfully navigates randomized stepping-stone fields.

Maintains stability under external pushes.

Abstract

Many real-world walking scenarios contain obstacles and unsafe ground patches (e.g., slippery or cluttered areas), leaving a disconnected set of admissible footholds that can be modeled as stepping-stone-like regions. We propose an onboard, perceptive mixed-integer model predictive control framework that jointly plans foot placement and step duration using step-to-step Divergent Component of Motion (DCM) dynamics. Ego-centric depth images are fused into a probabilistic local heightmap, from which we extract a union of convex steppable regions. Region membership is enforced with binary variables in a mixed-integer quadratic program (MIQP). To keep the optimization tractable while certifying safety, we embed capturability bounds in the DCM space: a lateral one-step condition (preventing leg crossing) and a sagittal infinite-step bound that limits unstable growth. We further re-plan within the step by back-propagating the measured instantaneous DCM to update the initial DCM, improving robustness to model mismatch and external disturbances. We evaluate the approach in simulation on Digit on randomized stepping-stone fields, including external pushes. The planner generates terrain-aware, dynamically consistent footstep sequences with adaptive timing and millisecond-level solve times.