ACID: Action Consistency via Inverse Dynamics for Planning with World Models
2026-07-02 • Robotics
RoboticsArtificial IntelligenceComputer Vision and Pattern Recognition
AI summaryⓘ
The authors point out that when robots plan actions using models that predict outcomes, they usually only check if the final predicted state is close to the goal, ignoring whether the steps taken to get there are actually doable. To fix this, they introduce ACID, a method that makes sure each predicted action can be matched by an inverse process, ensuring each step is realistic. They add this check into the planning cost, which helps plan better while using less computing power. Their method works well across various robot tasks and models.
decision-time planningaction-conditioned world modelsinverse dynamics modeltrajectory realismplanning costrobotic manipulationvisual navigationcycle consistencyembodied controlmodel-based planning
Authors
Gawon Seo, Dongwon Kim, Suha Kwak
Abstract
Decision-time planning with action-conditioned world models has become a popular paradigm for embodied control. However, the standard planning cost judges a candidate solely by how close its predicted terminal state lies to the goal, leaving the realizability of the intermediate transitions unchecked -- a predicted trajectory can look convincing while the environment rollout drifts away from it. In this paper, we propose ACID, a decision-time planning framework that introduces cycle action consistency: the action inferred backward from a predicted transition by an inverse dynamics model should recover the one that was conditioned on. We fold this per-step residual into the planning cost via a scale-invariant adaptive weight. Across four action-conditioned world models and six tasks spanning rigid and deformable manipulation, articulated control, and visual navigation, ACID consistently improves planning and matches the baseline's accuracy with substantially less planning compute.