Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Intention-aware Denoising Diffusion Model for Trajectory Prediction

Chen Liu, Shibo He, Haoyu Liu, Jiming Chen

TL;DR

An Intention-aware denoising Diffusion Model (IDM) is proposed, which decouple the original uncertainty into intention uncertainty and action uncertainty and model them with two dependent diffusion processes to decrease the inference time and reveal that introducing intention information is beneficial in modeling the diffusion process of fewer steps.

Abstract

Trajectory prediction is an essential component in autonomous driving, particularly for collision avoidance systems. Considering the inherent uncertainty of the task, numerous studies have utilized generative models to produce multiple plausible future trajectories for each agent. However, most of them suffer from restricted representation ability or unstable training issues. To overcome these limitations, we propose utilizing the diffusion model to generate the distribution of future trajectories. Two cruxes are to be settled to realize such an idea. First, the diversity of intention is intertwined with the uncertain surroundings, making the true distribution hard to parameterize. Second, the diffusion process is time-consuming during the inference phase, rendering it unrealistic to implement in a real-time driving system. We propose an Intention-aware denoising Diffusion Model (IDM), which tackles the above two problems. We decouple the original uncertainty into intention uncertainty and action uncertainty and model them with two dependent diffusion processes. To decrease the inference time, we reduce the variable dimensions in the intention-aware diffusion process and restrict the initial distribution of the action-aware diffusion process, which leads to fewer diffusion steps. To validate our approach, we conduct experiments on the Stanford Drone Dataset (SDD) and ETH/UCY dataset. Our methods achieve state-of-the-art results, with an FDE of 13.83 pixels on the SDD dataset and 0.36 meters on the ETH/UCY dataset. Compared with the original diffusion model, IDM reduces inference time by two-thirds. Interestingly, our experiments further reveal that introducing intention information is beneficial in modeling the diffusion process of fewer steps.

Intention-aware Denoising Diffusion Model for Trajectory Prediction

TL;DR

An Intention-aware denoising Diffusion Model (IDM) is proposed, which decouple the original uncertainty into intention uncertainty and action uncertainty and model them with two dependent diffusion processes to decrease the inference time and reveal that introducing intention information is beneficial in modeling the diffusion process of fewer steps.

Abstract

Trajectory prediction is an essential component in autonomous driving, particularly for collision avoidance systems. Considering the inherent uncertainty of the task, numerous studies have utilized generative models to produce multiple plausible future trajectories for each agent. However, most of them suffer from restricted representation ability or unstable training issues. To overcome these limitations, we propose utilizing the diffusion model to generate the distribution of future trajectories. Two cruxes are to be settled to realize such an idea. First, the diversity of intention is intertwined with the uncertain surroundings, making the true distribution hard to parameterize. Second, the diffusion process is time-consuming during the inference phase, rendering it unrealistic to implement in a real-time driving system. We propose an Intention-aware denoising Diffusion Model (IDM), which tackles the above two problems. We decouple the original uncertainty into intention uncertainty and action uncertainty and model them with two dependent diffusion processes. To decrease the inference time, we reduce the variable dimensions in the intention-aware diffusion process and restrict the initial distribution of the action-aware diffusion process, which leads to fewer diffusion steps. To validate our approach, we conduct experiments on the Stanford Drone Dataset (SDD) and ETH/UCY dataset. Our methods achieve state-of-the-art results, with an FDE of 13.83 pixels on the SDD dataset and 0.36 meters on the ETH/UCY dataset. Compared with the original diffusion model, IDM reduces inference time by two-thirds. Interestingly, our experiments further reveal that introducing intention information is beneficial in modeling the diffusion process of fewer steps.
Paper Structure (33 sections, 31 equations, 9 figures, 5 tables, 2 algorithms)

This paper contains 33 sections, 31 equations, 9 figures, 5 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Trajectory Prediction
  4. Physics-based approaches
  5. Planning-based approaches
  6. Pattern-based approaches
  7. Diffusion Model
  8. System Model
  9. Problem Statement
  10. Denoising Diffusion Probabilistic Model
  11. Conditional Denoising Diffusion Probabilistic Model
  12. Model description
  13. Architecture
  14. Goal Diffusion Process
  15. Trajectory Diffusion Process
  16. ...and 18 more sections

Figures (9)

  • Figure 1: Two kinds of uncertainty: (a) A pedestrian may turn left, turn right, or proceed straight based on his own will; (b) A pedestrian with a deterministic goal can also choose different paths in order to avoid collisions with their surroundings.
  • Figure 2: Denoising Diffusion Probabilistic Model
  • Figure 3: Overall structure of intention-aware trajectory diffusion model: a) EndNet: It models the diffusion process of the agent's endpoints. b) PathNet: It models the diffusion process of the agent's trajectories conditioned on a specific endpoint. c) PriorNet: It estimates the initial noise distribution of the trajectory diffusion process with fewer steps.
  • Figure 4: Prediction on SDD
  • Figure 5: Prediction on ETH/UCY
  • ...and 4 more figures