Core Concepts
Utilizing an Intention-aware Denoising Diffusion Model (IDM) can improve trajectory prediction accuracy in autonomous driving systems by modeling uncertainty and reducing inference time.
Abstract
The article introduces the IDM, a diffusion model that addresses uncertainty in trajectory prediction by decoupling intention and action uncertainties. The IDM consists of two processes: Goal Diffusion to model endpoint uncertainty and Trajectory Diffusion to model action uncertainty. By utilizing neural networks like EndNet, PriorNet, and PathNet, IDM efficiently generates future trajectories. Experiments on SDD and ETH/UCY datasets show state-of-the-art results with reduced inference time.
Stats
Our methods achieve state-of-the-art results, with an FDE of 13.83 pixels on SDD dataset and 0.36 meters on ETH/UCY datasets.
IDM reduces inference time by two-thirds compared to the original diffusion model.
Quotes
"Applying the original diffusion model would neglect the multimodal property of the trajectory."
"Our experiments further reveal that introducing intention information is beneficial in modeling the diffusion process of fewer steps."