Hierarchical World Models with Implicit Dynamics
Hierarchical World Models with Implicit Dynamics
This research introduces ID-WM (Implicit Dynamics World Models), a hierarchical framework that separates high-level state-space reachability sampling from low-level inverse dynamics solving. By prioritizing local physical feasibility, the method synthesizes novel long-horizon plans from short segments, achieving a 38.5% improvement over state-of-the-art policy and diffusion-based planners and a 135% gain over single-level world model MPC in complex, unseen configurations for navigation and manipulation tasks.
The paper proposes ID-WM, a hierarchical world model that decouples state-space reachability from local control to improve long-horizon planning for embodied agents.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
nstrained by low-level execution details. By prioritizing local physical feasibility over global behavioral patterns, ID-WM synthesizes novel long-horizon plans from short segments, transcending the specific behavioral modes seen during training. We demonstrate its effectiveness across three navigation and manipulation suites, achieving a 38.5% improvement over state-of-the-art policy and diffusion-based planners. Cr
Open source S4
. We introduce ID-WM (Implicit Dynamics World Models), a hierarchical world model that decouples state-space reachability from local control. Our framework utilizes a high-level model to sample feasible transitions across the state manifold and a low-level world model as an inverse dynamics solver to ground these transitions into precise actions. This allows the planner to identify reachable subgoals without being co
Open source S3
Source：arXiv Robotics — research abstracts · arxiv.org