Reproduce “Probabilistically-Safe Bipedal Navigation over Uncertain Terrain via Conformal Prediction and Contraction Analysis” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Kasidit Muenprasitivej, Ye Zhao, Glen Chou

Researchers at Georgia Institute of Technology developed a framework for probabilistically safe bipedal navigation over uncertain terrain by integrating Conformal Prediction for calibrated uncertainty quantification, an uncertainty-informed Model Predictive Control, and a contraction-based controller with flywheel torque. This approach enables bipedal robots to safely traverse challenging environments while maintaining high per-step invariance probabilities, such as 84.94% at an 85% confidence level on rough terrains.

Key insights

  • Conformal Prediction effectively provides empirically calibrated and significantly tighter uncertainty bounds for terrain elevation, allowing for less conservative yet statistically reliable planning decisions.
  • Integrating probabilistic safety directly into the high-level MPC through CP-informed constraints allows the robot to make more flexible, 'risk-seeking' navigation choices over uncertain regions with quantifiable confidence.
  • Control Contraction Metrics combined with RCI tubes and saltation matrices provide a formal method for guaranteeing robust tracking and forward invariance of the robot's centroidal states, bridging the gap between simplified planning and full-order dynamics under uncertainty.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.