Deep Active Inference with Diffusion Policy and Multiple Timescale World Model for Real-World Exploration and Navigation

arXiv 25.10 2025 55.4 method, application

TLDR

Deep active inference with diffusion policy and multiple timescale world model for real-world robotic exploration and navigation.

Reasoning

The paper presents a novel integration of diffusion policy and a multiple timescale recurrent state-space world model within an active inference framework, validated through real-world navigation experiments. Strengths include real-world empirical evaluation and a unified approach to exploration and goal-directed navigation; weaknesses are limited detail on limitations and comparisons in the abstract.

Read-first score

Read-first score 55.4, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 39.

Recency 8%

86.7

Uses a gentle age decay so recent papers surface without erasing older foundations. 2025

Methodology quality 25%

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

Topical relevance 42%

55.7

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

Reproducibility 25%

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

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 285.

Keyword Scores

world model

world dynamics prediction

model-based reinforcement learning world model

generative world model

world simulator

interactive world model

video world model

Deep Analysis

Innovations

Integration of diffusion policy as the policy model within a deep active inference framework
Multiple timescale recurrent state-space model (MTRSSM) as the world model for long-horizon latent imagination
Unified framework for exploration and goal-directed navigation via expected free energy minimization in real-world robotics

Methodology

The proposed deep active inference framework integrates a diffusion policy to generate diverse candidate actions and a multiple timescale recurrent state-space model (MTRSSM) to predict long-horizon consequences through latent imagination. Action selection is performed by minimizing the expected free energy (EFE), combining epistemic and extrinsic values. The framework is evaluated in real-world navigation experiments against baselines.

Key Results

The framework achieved higher success rates and fewer collisions compared with baselines, particularly in exploration-demanding scenarios.

Resources

Paper