PUMA: Perception-driven Unified Foothold Prior for Mobility Augmented Quadruped Parkour

2026-01-22

AI Summary

This work addresses the limitations of current quadrupedal robots in parkour tasks, which rely on predefined footholds and lack human-like environmental perception and real-time foothold decision-making, thereby constraining the exploration efficiency and adaptability of reinforcement learning. The authors propose an end-to-end learning framework that, for the first time, directly embeds a vision-driven polar-coordinate foothold prior into the policy network, unifying terrain perception and locomotion control within a single-stage training process. By eliminating the rigid foothold constraints inherent in conventional hierarchical architectures, the approach enables the robot to actively adjust its posture to navigate complex, discrete terrains. Experiments demonstrate that the system achieves exceptional agility and robustness in both simulation and real-world environments, significantly outperforming baseline methods that depend on precomputed footholds.

Technology Category

• Intelligent Robots: Learning & Optimization for ROB
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Search and Optimization: Learning to Search

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Parkour tasks for quadrupeds have emerged as a promising benchmark for agile locomotion. While human athletes can effectively perceive environmental characteristics to select appropriate footholds for obstacle traversal, endowing legged robots with similar perceptual reasoning remains a significant challenge. Existing methods often rely on hierarchical controllers that follow pre-computed footholds, thereby constraining the robot's real-time adaptability and the exploratory potential of reinforcement learning. To overcome these challenges, we present PUMA, an end-to-end learning framework that integrates visual perception and foothold priors into a single-stage training process. This approach leverages terrain features to estimate egocentric polar foothold priors, composed of relative distance and heading, guiding the robot in active posture adaptation for parkour tasks. Extensive experiments conducted in simulation and real-world environments across various discrete complex terrains, demonstrate PUMA's exceptional agility and robustness in challenging scenarios.

Problem

• quadruped locomotion
• perception-driven foothold selection
• agile parkour
• terrain adaptation
• real-time adaptability

Innovation

• end-to-end learning
• perception-driven foothold prior
• quadruped parkour
• terrain-aware locomotion
• egocentric polar representation