Introduces training-free and model-free trajectory planning by computing diffusion score functions directly from data libraries via kernel-weighted estimation.
April 2, 2026
Original Paper
Behavioral Score Diffusion: Model-Free Trajectory Planning via Kernel-Based Score Estimation from Data
arXiv · 2604.00391
The Takeaway
Moves away from learned parametric policies or analytical dynamics models. It achieves 98.5% of model-based performance using only 1,000 pre-collected trajectories, offering a highly scalable alternative for complex robotic systems.
From the abstract
Diffusion-based trajectory optimization has emerged as a powerful planning paradigm, but existing methods require either learned score networks trained on large datasets or analytical dynamics models for score computation. We introduce \emph{Behavioral Score Diffusion} (BSD), a training-free and model-free trajectory planner that computes the diffusion score function directly from a library of trajectory data via kernel-weighted estimation. At each denoising step, BSD retrieves relevant trajecto