Saved in:
Bibliographic Details
Main Authors: Hwang, Matthew, Liu, Yubin, Hakoda, Ryo, Oishi, Takeshi
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2604.02744
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908934683492352
author Hwang, Matthew
Liu, Yubin
Hakoda, Ryo
Oishi, Takeshi
author_facet Hwang, Matthew
Liu, Yubin
Hakoda, Ryo
Oishi, Takeshi
contents Quadrupedal locomotion over complex terrain has been a long-standing research topic in robotics. While recent reinforcement learning-based locomotion methods improve generalizability and foot-placement precision, they rely on implicit inference of foot positions from joint angles, lacking the explicit precision and stability guarantees of optimization-based approaches. To address this, we introduce a foot position map integrated into the heightmap, and a dynamic locomotion-stability reward within an attention-based framework to achieve locomotion on complex terrain. We validate our method extensively on terrains seen during training as well as out-of-domain (OOD) terrains. Our results demonstrate that the proposed method enables precise and stable movement, resulting in improved locomotion success rates on both in-domain and OOD terrains.
format Preprint
id arxiv_https___arxiv_org_abs_2604_02744
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Learning Locomotion on Complex Terrain for Quadrupedal Robots with Foot Position Maps and Stability Rewards
Hwang, Matthew
Liu, Yubin
Hakoda, Ryo
Oishi, Takeshi
Robotics
Quadrupedal locomotion over complex terrain has been a long-standing research topic in robotics. While recent reinforcement learning-based locomotion methods improve generalizability and foot-placement precision, they rely on implicit inference of foot positions from joint angles, lacking the explicit precision and stability guarantees of optimization-based approaches. To address this, we introduce a foot position map integrated into the heightmap, and a dynamic locomotion-stability reward within an attention-based framework to achieve locomotion on complex terrain. We validate our method extensively on terrains seen during training as well as out-of-domain (OOD) terrains. Our results demonstrate that the proposed method enables precise and stable movement, resulting in improved locomotion success rates on both in-domain and OOD terrains.
title Learning Locomotion on Complex Terrain for Quadrupedal Robots with Foot Position Maps and Stability Rewards
topic Robotics
url https://arxiv.org/abs/2604.02744