Safety-Critical Whole-Body Control for Humanoid Robots via Input-to-State Safe Control Barrier Functions

Kwanwoo Lee1, Sanghyuk Park1, Gyeongjae Park1, Myeong-Ju Kim2, Jaeheung Park1
1Dynamic Robotic Systems Lab, Seoul National University, Seoul, Republic of Korea
2Robotics Lab, Hyundai Motor Group, Uiwang, Republic of Korea

Abstract

Safety-critical control is essential for humanoid robots operating in complex human-centered environments that require safe physical interaction. Existing approaches remain limited because kinematic safety controllers do not explicitly account for full-body dynamics or contact stability, while dynamic whole-body controllers may fail to preserve kinematic safety guarantees under tracking errors and model mismatches. This paper presents a hierarchical safety-critical whole-body control framework for humanoid robots based on input-to-state safe control barrier functions (ISSf-CBFs). The proposed architecture integrates a kinematic-level whole-body controller (KinWBC), an ISSf-CBF safety filter, and a dynamic-level whole-body controller (DynWBC). KinWBC generates joint-motion references from prioritized tasks; the ISSf-CBF filter minimally modifies these references to satisfy kinematic safety constraints under bounded disturbances, and DynWBC tracks the modified references while enforcing full-body dynamics and contact stability. To transfer kinematic safety guarantees to full-order humanoid dynamics, the proposed framework formulates safety constraints on a whole-body kinematic model and tunes ISSf-CBF parameters to account for tracking errors and model mismatches. The framework handles multiple safety constraints, including joint limits, self-collision avoidance, obstacle avoidance, and workspace boundaries. Simulation and real-robot experiments demonstrate that the proposed framework improves safety margins under model mismatch and reliably enforces multiple safety constraints during manipulation, locomotion, and teleoperation.

Overview

Contributions

  • A hierarchical safety-critical whole-body control framework that accounts for kinematic safety, dynamic feasibility, and contact stability constraints.
  • A safety-transfer strategy from velocity kinematics to full-order dynamics that combines an ISSf-CBF-based safety filter with conservative parameter tuning, thereby preserving safety guarantees under bounded tracking errors and model uncertainties.
  • Validation through simulation and real-world humanoid locomotion, manipulation, and teleoperation experiments, demonstrating reliable safety enforcement under model mismatch and real-time operation across diverse scenarios.

Video Results

Simulation, teleoperation, and real-world obstacle-avoidance demonstrations.

Real Robot (Obstacle Avoidance 1)

Real Robot (Obstacle Avoidance 2)

Real Robot (Teleoperation)

Real Robot (Taichi motion)