⌁ Awesome World Model Hub Papers · Datasets · Projects
中文 GitHub
← Back to papers

Sampling-based Kinodynamic Planning with World Models

arXiv 2026 33.6 method

TLDR

This paper integrates kinodynamic sampling-based planning with visual world models, using latent-space RRT/KPIECE to outperform CEM/MPPI in long-horizon tasks.

Reasoning

The paper addresses a clear limitation of optimizer-based planning with world models and proposes GPU-native sampling-based alternatives, with experiments in simulation and on a real robot. The main weakness visible from the abstract is the narrow evaluation setting and limited comparison to other planning baselines.

Read-first score

Read-first score 33.6, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 44.

Recency 5%
100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Topical relevance 25%
62.9

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Methodology quality 15%
50

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=evaluation

Reproducibility 15%
30

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Graph prestige 15%
5.7

PageRank-style prestige over the local citation graph built from available OpenAlex references. local_pagerank=0.0016

Citation impact 15%
0

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. cited_by_count=0

Citation velocity 10%
0

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 337.

Keyword Scores

world model
10
world simulator
7
world dynamics prediction
7
interactive world model
6
generative world model
5
video world model
5
model-based reinforcement learning world model
4

Deep Analysis

Innovations

  • Revisits kinodynamic sampling-based motion planning for visual world models by using the learned model as the propagator of a tree search in latent space.
  • Shows that sampling strategies from high-dimensional configuration-space planning apply directly to world-model latent spaces, enabling expansion through intermediate states that myopic optimizers miss.
  • Introduces two latent-space exploration strategies: nearest-neighbor-guided RRT and projection-based KPIECE.
  • Uses GPU-native batched expansion for efficient device utilization during planning.

Methodology

The method uses an off-the-shelf DINO-WM visual world model as the propagator for sampling-based tree search in latent space. It adapts kinodynamic planning strategies, specifically nearest-neighbor-guided RRT and projection-based KPIECE, with GPU-native batched expansion. Evaluation compares against CEM and MPPI on four simulated domains and a real-world block-pushing task using success rate and planning time.

Key Results

The proposed approaches match or exceed the success rate of CEM and MPPI at up to 13.8x lower planning time in simulation, and improve success rate by 20% on a real robot block-pushing task.

Resources

Paper

Tags