Driving · 2026-10-06

Tackling Sim-to-Real Mismatch Through Sampling-Based Disturbance Observers: From Analytical Models to Learned World Models

Tianqi Zhu, Jun Yang, Jianliang Mao, Cong Li, Shihua Li

arXiv:2610.04896PDF

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

TL;DR

The authors introduce SDOB (sampling-based disturbance observer) to extend disturbance-observer compensation to modern predictive models accessed via forward rollouts or cost-query interfaces (including simulators and learned world models). SDOB separates mismatch into (1) a state-effect disturbance that changes rollout states and (2) a cost disturbance that changes queried costs at fixed query states, then estimates and compensates both online for improved control performance.

Why it matters

The paper targets an “interface mismatch”: classical DOB assumes explicit plant dynamics and a disturbance channel, while sampling-based MPC and learned world models provide only rollout/cost-query access. By defining disturbance in two observable channels and estimating them from prediction-vs-feedback mismatch, SDOB makes disturbance-aware feedback compensation applicable to analytical models, simulators, and learned world models.

Method

  • Introduce SDOB interface-level disturbance representation with two channels: state-effect disturbances (prediction vs feedback state mismatch) and cost disturbances (cost changes at fixed query states across perceived-environment updates).
  • Estimate state-effect disturbance online via sampling-based rollout scoring: draw candidate disturbance values, compute residuals between feedback and predictions, and fit using a weighted mean over candidates.
  • Estimate cost disturbance by matching local cost-field changes using repeated cost queries before/after environment updates, treating local motion in query space as a disturbance. Planning uses the estimated disturbances for compensation.

Limitation

Remark 1 states three conditions for SDOB: (1) a repeatable rollout interface or comparable cost queries across updates, (2) the selected disturbance coordinates must affect feedback or queried costs (otherwise unobservable), and (3) disturbances are assumed to vary slowly between successive updates; abrupt topology changes or several unresolved motions in the same neighborhood can violate the cost-transport approximation.

Abstract (from arXiv)

Robotic controllers increasingly rely on analytical models, simulators, cost-query interfaces, and learned world models. However, physical deployment can deviate from nominal assumptions, and additional disturbances may arise even when the model itself is accurate. In control systems, disturbance observers (DOB) are widely used to estimate such unmeasured effects from nominal models and measured feedback. Classical DOB formulations are generally built around explicit plant models. This paper develops the sampling-based disturbance observer (SDOB), extending the DOB principle to a broader range of models, including simulators and learned world models, through state-rollout or cost-query interfaces. SDOB separates two observable channels: state-effect disturbances, for which the feedback state differs from its prediction, and cost disturbances, for which the same query state receives different costs as the perceived environment changes. Diverse simulation and real-robot experiments across traditional and learned models demonstrate the effectiveness of SDOB in compensating for sim-to-real mismatch and improving control performance.

Related papers

Publication record

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

First public release: 2026-10-04 · Indexed by Pub-AI: 2026-10-06 UTC

preprint

Tackling Sim-to-Real Mismatch Through Sampling-Based Disturbance Observers: From Analytical Models to Learned World Models

Tianqi Zhu, Jun Yang, Jianliang Mao, Cong Li, Shihua Li

Preprint posted: 2026-10-04 · Original record ↗

arxiv
2610.04896
Explicit source versions (1)
  • v1 · 2026-10-04
  1. Preprint available

    Source date: 2026-10-04 · First detected: 2026-10-25 UTC · Historical baseline

    Source evidence (1)
Abstract · arxiv

Robotic controllers increasingly rely on analytical models, simulators, cost-query interfaces, and learned world models. However, physical deployment can deviate from nominal assumptions, and additional disturbances may arise even when the model itself is accurate. In control systems, disturbance observers (DOB) are widely used to estimate such unmeasured effects from nominal models and measured feedback. Classical DOB formulations are generally built around explicit plant models. This paper develops the sampling-based disturbance observer (SDOB), extending the DOB principle to a broader range of models, including simulators and learned world models, through state-rollout or cost-query interfaces. SDOB separates two observable channels: state-effect disturbances, for which the feedback state differs from its prediction, and cost disturbances, for which the same query state receives different costs as the perceived environment changes. Diverse simulation and real-robot experiments across traditional and learned models demonstrate the effectiveness of SDOB in compensating for sim-to-real mismatch and improving control performance.