Driving · 2026-10-01

LocoWM: High-Precision Locomotion through World-Model-Guided Residual Adaptation

Zijie Zhao, Shengqian Chen, Xiaoxu Wang, Han Jiang, Yuanheng Zhu, Dongbin Zhao

AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.

TL;DR

LocoWM is a world-model-guided residual adaptation framework for high-precision locomotion. A base policy provides command-following actions, and an action-conditioned world model predicts future task substates from proprioceptive history and the base action. A residual adapter uses the predicted sequence to output additive action corrections. Two-stage training freezes locomotion and dynamics before training the adapter for precision.

Why it matters

The authors target high-precision locomotion where command tracking must coexist with precise regulation of task-relevant physical states. They argue end-to-end training can under-optimize precision and that reactive residual control corrects only after deviations become observable. By forecasting task substates and using them to condition residual corrections in a single forward pass, LocoWM aims to make residual control preactive rather than reactive.

Method

Base policy proposes actions for command-following; a learned, action-conditioned world model predicts a future sequence of task substates from proprioceptive history and the base action.
A residual adapter conditions on the predicted task-substate sequence to generate additive action corrections (executed action = base action + residual).
Two-stage training: learn locomotion and action-conditioned dynamics first, then freeze them and train only the residual adapter with the full reward.

Abstract (from arXiv)

High-precision locomotion combines motion-command tracking with precise regulation of task-relevant physical states, enabling robots to interact reliably with their surroundings during motion. Joint end-to-end optimization can leave precision objectives insufficiently optimized, while reactive residual control adjusts actions only after deviations become observable. We present \textbf{LocoWM}, a world-model-guided preactive residual adaptation framework for high-precision locomotion. A base policy provides command-following locomotion, while an action-conditioned world model predicts a sequence of future physical states from proprioceptive history and the proposed base action. A residual adapter conditions on this predicted sequence to generate additive action corrections that compensate for anticipated deviations. Two-stage training first learns locomotion and action-conditioned dynamics, then freezes both modules while training the adapter, separating locomotion acquisition from precision adaptation. Experiments spanning terrain leveling, acceleration compensation, and push recovery demonstrate improved control precision and disturbance robustness over end-to-end and reactive residual baselines. Demos and code are available at: https://zhaozijie2022.github.io/LocoWM

Publication record

Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.

First public release: 2026-09-30 · Indexed by Pub-AI: 2026-10-01 UTC

preprint