Driving · 2026-10-02

Training-Free Diffusion Planning with Analytical Local Scores

Michael Y. Fatemi, Jinhao Liang, Ferdinando Fioretto

arXiv:2610.01959PDF

AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.

TL;DR

The authors propose TFDP, a training-free diffusion-based motion planner. Instead of learning a global neural score from large feasible trajectory datasets, TFDP builds an analytical surrogate score from local smoothness, obstacle avoidance, inter-agent separation, and velocity-feasibility terms. It refines a coarse trajectory using an iterative denoising procedure with a decreasing noise schedule, scaling to large multi-agent scenes.

Why it matters

Diffusion planners for motion planning usually require learning a neural score and thus depend on large, feasible trajectory datasets that are map/distribution specific. TFDP aims to keep diffusion-like iterative refinement while avoiding training and explicitly tying denoising updates to motion planning constraints, including scalability to 300-agent/100-obstacle settings with limited compute.

Method

  • Builds a training-free surrogate trajectory distribution as a product of smoothness and feasibility indicators (obstacle avoidance, inter-agent separation, and kinematic bound), then uses analytical local score terms for denoising updates.
  • Approximates the full noisy trajectory score using a local score decomposition dominated by neighboring waypoint interactions and nearby constraints (smoothness, obstacles, and same-time agent separation).
  • Refines an initial coarse trajectory with iterative denoising under a decreasing noise schedule, while enforcing velocity limits via a soft penalty and using analytical constraint-score terms tied to feasibility probabilities.

Limitation

The method relies on approximating the full trajectory score with a local score decomposition (computed from local neighborhoods); the local independence structure is approximately preserved only in the low-noise regime, motivating short score windows at low noise levels.

Abstract (from arXiv)

Path finding and multi-robot motion planning require trajectories that are smooth, goal-directed, and collision-free in environments with complex geometric constraints. Recent diffusion-based planners have shown that trajectory generation can be cast as iterative denoising which has opened the doors to learning-based approaches that can handle multi-modal trajectory distributions and refine entire trajectories. However, a key limitation is that diffusion planners require training on large collections of feasible trajectories, rendering them map-specific, and difficult to deploy when high-quality demonstrations are unavailable. This paper introduces a training-free diffusion-based motion planner that replaces learned global trajectory scores with analytical local scores derived from obstacle, smoothness, velocity, and inter-agent feasibility terms. The proposed idea relies on a key observation: the score of a trajectory can be reconstructed by considering only local interactions between neighboring waypoints and nearby constraints. This structure exploitation yields a decomposed denoising procedure that retains the optimization structure of classical trajectory methods while inheriting the iterative refinement behavior of diffusion models. Experiments on a large collection of complex environments and large multi-agent planning tasks show that the proposed analytical score produces smooth and feasible trajectories within limited computational costs, for example in generating feasible paths for 300+ agents in environments containing 100+ obstacles in under 6 seconds on a GPU, outperforming strong learning-based and optimization baselines, while avoiding the data requirements of learned diffusion planners.

Related papers

Publication record

Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.

First public release: 2026-10-01 · Indexed by Pub-AI: 2026-10-02 UTC

preprint

Training-Free Diffusion Planning with Analytical Local Scores

Michael Y. Fatemi, Jinhao Liang, Ferdinando Fioretto

Preprint posted: 2026-10-01 · Original record ↗

arxiv
2610.01959
Explicit source versions (1)
  • v1 · 2026-10-01
  1. Preprint available

    Source date: 2026-10-01 · First detected: 2026-10-25 UTC · Historical baseline

    Source evidence (1)
Abstract · arxiv

Path finding and multi-robot motion planning require trajectories that are smooth, goal-directed, and collision-free in environments with complex geometric constraints. Recent diffusion-based planners have shown that trajectory generation can be cast as iterative denoising which has opened the doors to learning-based approaches that can handle multi-modal trajectory distributions and refine entire trajectories. However, a key limitation is that diffusion planners require training on large collections of feasible trajectories, rendering them map-specific, and difficult to deploy when high-quality demonstrations are unavailable. This paper introduces a training-free diffusion-based motion planner that replaces learned global trajectory scores with analytical local scores derived from obstacle, smoothness, velocity, and inter-agent feasibility terms. The proposed idea relies on a key observation: the score of a trajectory can be reconstructed by considering only local interactions between neighboring waypoints and nearby constraints. This structure exploitation yields a decomposed denoising procedure that retains the optimization structure of classical trajectory methods while inheriting the iterative refinement behavior of diffusion models. Experiments on a large collection of complex environments and large multi-agent planning tasks show that the proposed analytical score produces smooth and feasible trajectories within limited computational costs, for example in generating feasible paths for 300+ agents in environments containing 100+ obstacles in under 6 seconds on a GPU, outperforming strong learning-based and optimization baselines, while avoiding the data requirements of learned diffusion planners.