ACID: Action Consistency via Inverse Dynamics for Planning with World Models

2026-07-02

AI Summary

This work addresses the discrepancy between simulation and reality in world model–based decision planning, which arises when intermediate state transitions are infeasible. To mitigate this issue, the authors propose the ACID framework, which introduces an inverse dynamics model during planning and establishes a novel cyclic action consistency mechanism. This mechanism enhances planning reliability by verifying the feasibility of actions at each step of the predicted trajectory. Furthermore, the framework incorporates a scale-invariant adaptive weighting scheme that integrates consistency residuals into the planning cost function. Experimental results demonstrate that ACID significantly outperforms baseline methods across four distinct world models and six benchmark tasks, achieving comparable or superior planning accuracy with lower computational overhead.

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.

Problem

• decision-time planning
• action-conditioned world models
• trajectory realizability
• inverse dynamics
• planning consistency

Innovation

• inverse dynamics
• action consistency
• decision-time planning
• world models
• cycle consistency