Extendable Long-Horizon Planning via Hierarchical Multiscale Diffusion

2025-03-25

AI Summary

This work addresses the error accumulation problem encountered by agents when generating out-of-distribution, ultra-long temporal trajectories. To this end, we propose a novel scalable long-horizon motion planning paradigm. Methodologically, we introduce the Hierarchical Multi-scale Diffuser (HM-Diffuser), the first diffusion-based architecture featuring progressive trajectory expansion (PTE) and adaptive planning contemplation, augmented with a recursive HM-Diffuser structure to enable dynamic-length planning. By incorporating hierarchical temporal modeling, multi-scale feature fusion, and trajectory stitching enhancement, our approach significantly improves generalization to trajectories substantially longer than those seen during training. Empirical evaluation across multiple robotic simulation tasks demonstrates stable generation of high-fidelity motion sequences up to several times longer than training horizons. This advances diffusion-based planners toward practical deployment in long-horizon autonomous navigation and manipulation scenarios.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Multiagent Systems: Multiagent Planning
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

This paper tackles a novel problem, extendable long-horizon planning-enabling agents to plan trajectories longer than those in training data without compounding errors. To tackle this, we propose the Hierarchical Multiscale Diffuser (HM-Diffuser) and Progressive Trajectory Extension (PTE), an augmentation method that iteratively generates longer trajectories by stitching shorter ones. HM-Diffuser trains on these extended trajectories using a hierarchical structure, efficiently handling tasks across multiple temporal scales. Additionally, we introduce Adaptive Plan Pondering and the Recursive HM-Diffuser, which consolidate hierarchical layers into a single model to process temporal scales recursively. Experimental results demonstrate the effectiveness of our approach, advancing diffusion-based planners for scalable long-horizon planning.

Problem

• Enabling agents to plan beyond training data horizons
• Reducing compounding errors in long trajectory planning
• Handling multi-scale tasks with hierarchical diffusion models

Innovation

• Hierarchical Multiscale Diffuser for scalable planning
• Progressive Trajectory Extension via stitching
• Recursive HM-Diffuser consolidates temporal scales