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

2025-10-27

AI Summary

This work addresses the challenge of simultaneous environment exploration and goal-directed navigation for autonomous robots in real-world scenarios. We propose a unified deep active inference framework that drives intelligent decision-making by minimizing expected free energy. Our key contribution is the first integration of diffusion-based policies with a multi-timescale recurrent state-space model (MTRSSM) within an active inference architecture—enabling long-horizon future imagination in latent space, diverse action sampling, and adaptive exploration–exploitation trade-offs. Experiments on physical robot navigation tasks demonstrate significant improvements in task success rate and substantial reductions in collision frequency, particularly under high environmental uncertainty and demanding exploration conditions.

Technology Category

• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Model-Based Reasoning

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

Autonomous robotic navigation in real-world environments requires exploration to acquire environmental information as well as goal-directed navigation in order to reach specified targets. Active inference (AIF) based on the free-energy principle provides a unified framework for these behaviors by minimizing the expected free energy (EFE), thereby combining epistemic and extrinsic values. To realize this practically, we propose a deep AIF framework that integrates a diffusion policy as the policy model and a multiple timescale recurrent state-space model (MTRSSM) as the world model. The diffusion policy generates diverse candidate actions while the MTRSSM predicts their long-horizon consequences through latent imagination, enabling action selection that minimizes EFE. Real-world navigation experiments demonstrated that our framework achieved higher success rates and fewer collisions compared with the baselines, particularly in exploration-demanding scenarios. These results highlight how AIF based on EFE minimization can unify exploration and goal-directed navigation in real-world robotic settings.

Problem

• Unifying exploration and goal-directed navigation for robots
• Minimizing expected free energy using diffusion policies
• Improving real-world navigation success with collision reduction

Innovation

• Diffusion policy generates diverse candidate actions
• Multiple timescale world model predicts long-term consequences
• Deep active inference minimizes expected free energy