TrajFlow: Multi-modal Motion Prediction via Flow Matching

2025-06-10

AI Summary

To address the inefficiency and inadequate modeling of diversity and uncertainty in multimodal motion prediction for autonomous driving in dynamic scenarios, this paper proposes a single-forward-pass inference framework based on flow matching. Our key contributions are: (1) a novel single-pass multimodal trajectory generation mechanism that eliminates iterative sampling; (2) incorporation of a Plackett–Luce ranking loss to explicitly model confidence-based trajectory ordering, thereby improving uncertainty estimation quality; and (3) a self-conditioned training strategy that reuses intermediate predictions to construct noisy inputs, enhancing temporal consistency. Experiments on the Waymo Open Motion Dataset demonstrate state-of-the-art performance: a 4.2% reduction in final displacement error (FDE), a 23% improvement in trajectory diversity, and a 68% reduction in computational overhead—achieving a superior balance among accuracy, diversity, and inference efficiency.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Efficient and accurate motion prediction is crucial for ensuring safety and informed decision-making in autonomous driving, particularly under dynamic real-world conditions that necessitate multi-modal forecasts. We introduce TrajFlow, a novel flow matching-based motion prediction framework that addresses the scalability and efficiency challenges of existing generative trajectory prediction methods. Unlike conventional generative approaches that employ i.i.d. sampling and require multiple inference passes to capture diverse outcomes, TrajFlow predicts multiple plausible future trajectories in a single pass, significantly reducing computational overhead while maintaining coherence across predictions. Moreover, we propose a ranking loss based on the Plackett-Luce distribution to improve uncertainty estimation of predicted trajectories. Additionally, we design a self-conditioning training technique that reuses the model's own predictions to construct noisy inputs during a second forward pass, thereby improving generalization and accelerating inference. Extensive experiments on the large-scale Waymo Open Motion Dataset (WOMD) demonstrate that TrajFlow achieves state-of-the-art performance across various key metrics, underscoring its effectiveness for safety-critical autonomous driving applications. The code and other details are available on the project website https://traj-flow.github.io/.

Problem

• Efficient multi-modal motion prediction for autonomous driving
• Scalability and efficiency challenges in trajectory prediction
• Improved uncertainty estimation for predicted trajectories

Innovation

• Flow matching for multi-modal trajectory prediction
• Single-pass prediction reduces computational overhead
• Self-conditioning training improves generalization