UEREBot: Learning Safe Quadrupedal Locomotion under Unstructured Environments and High-Speed Dynamic Obstacles

TL;DR

UEREBot addresses safe quadrupedal locomotion in unstructured environments with uneven terrain and high-speed dynamic obstacles by separating slow planning from rapid reflexive evasion within a constrained OCP blueprint. It introduces a spatial--temporal planner that outputs a reference path, dynamic obstacle predictions, and a threat score; a threat-aware handoff that softly fuses navigation and reflex commands; and a CBF safety shield that enforces hard safety during execution. Validated in Isaac Lab simulation and on a Unitree Go2 with onboard perception, UEREBot shows improved avoidance success and larger safety margins while preserving goal progress, outperforming representative baselines; ablations confirm the importance of the handoff, CBF shield, and obstacle prediction. The work offers a practical, scalable approach to reliable quadrupedal locomotion in dynamic, unstructured environments, enabling more robust real-world deployment.

Abstract

Quadruped robots are increasingly deployed in unstructured environments. Safe locomotion in these settings requires long-horizon goal progress, passability over uneven terrain and static constraints, and collision avoidance against high-speed dynamic obstacles. A single system cannot fully satisfy all three objectives simultaneously: planning-based decisions can be too slow, while purely reactive decisions can sacrifice goal progress and passability. To resolve this conflict, we propose UEREBot (Unstructured-Environment Reflexive Evasion Robot), a hierarchical framework that separates slow planning from instantaneous reflexive evasion and coordinates them during execution. UEREBot formulates the task as a constrained optimal control problem blueprint. It adopts a spatial--temporal planner that provides reference guidance toward the goal and threat signals. It then uses a threat-aware handoff to fuse navigation and reflex actions into a nominal command, and a control barrier function shield as a final execution safeguard. We evaluate UEREBot in Isaac Lab simulation and deploy it on a Unitree Go2 quadruped equipped with onboard perception. Across diverse environments with complex static structure and high-speed dynamic obstacles, UEREBot achieves higher avoidance success and more stable locomotion while maintaining goal progress than representative baselines, demonstrating improved safety--progress trade-offs.

Figures (13)

• Figure 1: Overview of UEREBot in complex dynamic environments. The robot tracks a reference path toward the goal region while responding to dynamic attacks under limited reaction time. Green arrows indicate the reference direction, red arrows indicate the attack direction of the dynamic obstacle, and yellow circle marks the goal location. (a) Indoor room. (b) Indoor corridor with steps. (c) Outdoor narrow path. (d) Outdoor grass. (e) Outdoor steps. The attacks include poke, hit, and kick, as well as a quadruped and a humanoid robot.
• Figure 2: Task overview. The robot tracks a reference path toward the goal region in a complex environment with rough terrain and static obstacles. Along the path, the robot may face dynamic attacks, including poke, hit, and kick.
• Figure 3: Overview of UEREBot framework. Perception and state estimation provide the robot state and obstacle observations, and an offline 2.5D map is queried to obtain the passability field for terrain and static obstacles. A spatial--temporal planner uses these inputs to produce a reference path, dynamic obstacle predictions, and a threat score. A navigation policy and a reflex policy then generate candidate commands, which are fused by a threat-aware handoff. The fused command is filtered by a CBF shield and executed by the low-level locomotion controller.
• Figure 4: Representative simulation experiments. The robot performs locomotion toward the goal region while traversing rough terrain, passing static obstacles in open and confined spaces, and avoiding high-speed dynamic obstacles under limited reaction time. (a) Open flat ground with static and dynamic obstacles. (b) Narrow corridor. (c) Rough terrain. (d) Stair traversal.
• Figure 5: Quantitative results comparing UEREBot against baselines in terms of overall success and its safety–progress trade-off, robustness under increasing attack speed, and the distribution of locomotion cost.