LoComposition: Terrain-Adaptive Energy-Efficient Quadruped Locomotion without Gait Priors

2026-06-14

AI Summary

Traditional quadrupedal locomotion control relies on complex reward functions that entangle task objectives, constraints, gait preferences, and terrain adaptation, leading to optimization challenges and poor energy efficiency. This work proposes a decoupled hierarchical reinforcement learning framework in which rewards encode task goals, constraints enforce operational limits, energy minimization guides the emergence of efficient gaits, and exteroceptive sensing—via LiDAR-based elevation maps—enables terrain adaptation without requiring predefined gait priors. The approach yields highly efficient locomotion policies that, compared to conventional methods, reduce transport cost by 56% and constraint violations by 96% while preserving terrain traversability. Furthermore, the learned policy demonstrates zero-shot transfer to a real-world Unitree Go2 robot, showcasing its robustness and practical applicability.

Abstract

Learning-based quadrupedal locomotion typically relies on complex reward formulations that entangle task specification, operational limits, gait preference, and terrain adaptation within a single optimization objective. We instead treat these functions through distinct mechanisms: rewards for task specification, constraints for operational limits, energy minimization for gait preference, and exteroceptive perception for adapting energy use to terrain difficulty. We show that these components jointly enable efficient, terrain-adaptive locomotion, and that removing each component exposes a distinct failure mode. Our formulation removes explicit gait priors (including air-time, contact-count, and foot-clearance targets) in favor of emergent behavior. Compared to a conventional complex-reward baseline, our formulation achieves comparable terrain traversal while reducing cost of transport by 56% and operational-limit violations by 96%. The resulting policies transfer zero-shot to a physical Unitree Go2 using LiDAR-based elevation mapping. Project website with videos: https://tinyurl.com/locomposition.

Problem

• quadruped locomotion
• energy efficiency
• terrain adaptation
• gait priors
• reward formulation

Innovation

• terrain-adaptive locomotion
• energy-efficient control
• gait-free learning
• constraint-based RL
• exteroceptive perception