TRANS: Terrain-aware Reinforcement Learning for Agile Navigation of Quadruped Robots under Social Interactions

arXiv cs.RO / 4/1/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper presents TRANS, a terrain-aware deep reinforcement learning framework for agile quadruped navigation in unstructured environments with social interactions.
It argues that existing quadruped approaches either decouple planning and locomotion (missing whole-body/terrain constraints) or rely on end-to-end sensing that is high-frequency, noisy, and computationally expensive.
TRANS uses a two-stage training setup with three DRL pipelines: TRANS-Loco for locomotion over uneven terrain without explicit terrain/contact observations, TRANS-Nav for social navigation using transformed LiDAR input under differential-drive kinematics, and a unified TRANS pipeline combining both.
Benchmarks against locomotion and social-navigation baselines show TRANS’s effectiveness, and hardware experiments indicate potential sim-to-real transfer.

Abstract

This study introduces TRANS: Terrain-aware Reinforcement learning for Agile Navigation under Social interactions, a deep reinforcement learning (DRL) framework for quadrupedal social navigation over unstructured terrains. Conventional quadrupedal navigation typically separates motion planning from locomotion control, neglecting whole-body constraints and terrain awareness. On the other hand, end-to-end methods are more integrated but require high-frequency sensing, which is often noisy and computationally costly. In addition, most existing approaches assume static environments, limiting their use in human-populated settings. To address these limitations, we propose a two-stage training framework with three DRL pipelines. (1) TRANS-Loco employs an asymmetric actor-critic (AC) model for quadrupedal locomotion, enabling traversal of uneven terrains without explicit terrain or contact observations. (2) TRANS-Nav applies a symmetric AC framework for social navigation, directly mapping transformed LiDAR data to ego-agent actions under differential-drive kinematics. (3) A unified pipeline, TRANS, integrates TRANS-Loco and TRANS-Nav, supporting terrain-aware quadrupedal navigation in uneven and socially interactive environments. Comprehensive benchmarks against locomotion and social navigation baselines demonstrate the effectiveness of TRANS. Hardware experiments further confirm its potential for sim-to-real transfer.