Safety-critical Locomotion of Biped Robots in Infeasible Paths: Overcoming Obstacles during Navigation toward Destination

TL;DR

The paper tackles safety-critical locomotion for legged robots facing infeasible paths due to obstacles by introducing a three-layer framework that first estimates dynamic properties from physical interactions, then applies a hierarchical control barrier-function–based planner to generate a safe velocity reference, and finally employs a disturbance-observer–augmented Hybrid Linear Inverted Pendulum (H-LIP) walking controller with a whole-body controller for robust task tracking. A mass-based safety hierarchy prioritizes avoiding heavier objects ($h_1 \gg h_2$) to resolve conflicts without explicit global planning, while a DOB compensates disturbances from object interactions to maintain stability. Key contributions include (1) autonomous object-property estimation guiding safety hierarchy, (2) a model-free safety-critical planning layer that yields safe motion without full motion planning, and (3) DOB-enhanced H-LIP walking that improves tracking and resilience during contact. Simulations with the Cassie robot demonstrate successful navigation by kicking lighter obstacles aside and reaching the goal with reduced yaw error, highlighting practical implications for safe navigation in cluttered, unstructured environments.

Abstract

This paper proposes a safety-critical locomotion control framework employed for legged robots exploring through infeasible path in obstacle-rich environments. Our research focus is on achieving safe and robust locomotion where robots confront unavoidable obstacles en route to their designated destination. Through the utilization of outcomes from physical interactions with unknown objects, we establish a hierarchy among the safety-critical conditions avoiding the obstacles. This hierarchy enables the generation of a safe reference trajectory that adeptly mitigates conflicts among safety conditions and reduce the risk while controlling the robot toward its destination without additional motion planning methods. In addition, robust bipedal locomotion is achieved by utilizing the Hybrid Linear Inverted Pendulum model, coupled with a disturbance observer addressing a disturbance from the physical interaction.

Figures (5)

• Figure 1: Safety-critical locomotion scenario: Our framework enables safe locomotion when the robot is blocked by unknown objects.
• Figure 2: Structure of the proposed framework: The first layer focuses on estimating the unknown dynamic properties of objects (Section \ref{['section3_1']}). With the estimated properties, a hierarchy among the safety conditions is established, and the safe velocity input is computed in the second layer (Section \ref{['section3_2']}). The last layer is dedicated to robust locomotion involving the use of H-LIP with DOB (Section \ref{['section3_3']}) and the WBC (Section \ref{['Section3_4']}).
• Figure 3: Disturbance Observer: A typical disturbance observer for a linear system is employed with "Q-filter" since the H-LIP is the nominal system.
• Figure 4: Simulation results: (a) snapshots of simulation demonstrating baseline locomotion control, (b) snapshot for the proposed locomotion control, (c) x and y positions of the floating base of the robot, (d) footstep of swing foot in the x direction, (e) yaw orientation of the floating base of the robot, (f) CBF values of safety-critical conditions. We implement three locomotion control methods: 1) baseline method (green), 2) CBF-based safe locomotion control (blue), 3) CBF-based safe locomotion control with DOB (red).
• Figure 5: Approximated estimation of mass: Three data sets (1000, 2000, 3000) are utilized to estimate the mass of each obstacle.

Theorems & Definitions (1)

• Definition 1