ADM: Accelerated Diffusion Model via Estimated Priors for Robust Motion Prediction under Uncertainties

TL;DR

This paper introduces a diffusion-based motion prediction framework that accelerates inference and improves robustness, enabling real-time trajectory prediction for autonomous driving by learning a coarse prior to skip denoising steps.

Contribution

It presents a novel accelerated diffusion model that learns a coarse prior to reduce denoising steps, enhancing speed and noise resistance for real-time multi-agent motion prediction.

Findings

Inference time reduced to 136ms

Significant improvement on Argoverse dataset

Enhanced noise robustness in predictions

Abstract

Motion prediction is a challenging problem in autonomous driving as it demands the system to comprehend stochastic dynamics and the multi-modal nature of real-world agent interactions. Diffusion models have recently risen to prominence, and have proven particularly effective in pedestrian motion prediction tasks. However, the significant time consumption and sensitivity to noise have limited the real-time predictive capability of diffusion models. In response to these impediments, we propose a novel diffusion-based, acceleratable framework that adeptly predicts future trajectories of agents with enhanced resistance to noise. The core idea of our model is to learn a coarse-grained prior distribution of trajectory, which can skip a large number of denoise steps. This advancement not only boosts sampling efficiency but also maintains the fidelity of prediction accuracy. Our method meets the rigorous real-time operational standards essential for autonomous vehicles, enabling prompt trajectory generation that is vital for secure and efficient navigation. Through extensive experiments, our method speeds up the inference time to 136ms compared to standard diffusion model, and achieves significant improvement in multi-agent motion prediction on the Argoverse 1 motion forecasting dataset.