Problem
Research questions and friction points this paper is trying to address.
Predicting human road user behavior variability
Capturing distributions over trajectory features
Improving probabilistic trajectory prediction accuracy
TrajFlow: Learning Distributions over Trajectories for Human Behavior Prediction
🤖 AI Summary
Accurately predicting the future trajectories of human road users is critical for risk-aware autonomous driving, yet the inherent uncertainty poses significant challenges for probabilistic trajectory distribution modeling. This paper proposes a novel probabilistic trajectory prediction framework that decouples distribution learning into an abstract latent feature space. By co-designing a variational autoencoder (VAE) and normalizing flows, the method enables end-to-end differentiable, highly expressive, and interpretable probabilistic generation. Crucially, it is the first to shift trajectory uncertainty modeling from the raw observation space to a structured latent space, substantially improving modeling fidelity and generalization. The approach achieves significant gains over state-of-the-art methods on two controllable synthetic benchmarks and attains leading or top-tier performance on real-world datasets—including ETH/UCY, rounD, and nuScenes.
📝 Abstract
Predicting the future behavior of human road users is an important aspect for the development of risk-aware autonomous vehicles. While many models have been developed towards this end, effectively capturing and predicting the variability inherent to human behavior still remains an open challenge. This paper proposes TrajFlow—a new approach for probabilistic trajectory prediction based on Normalizing Flows. We reformulate the problem of capturing distributions over trajectories into capturing distributions over abstracted trajectory features using an autoencoder, simplifying the learning task of the Normalizing Flows. TrajFlow outperforms state-of-the-art behavior prediction models in capturing full trajectory distributions in two synthetic benchmarks with known true distributions, and is competitive on the naturalistic datasets ETH/UCY, rounD, and nuScenes. Our results demonstrate the effectiveness of TrajFlow in probabilistic prediction of human behavior.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Normalizing Flows for trajectory prediction
Autoencoder abstracts trajectory features
Outperforms state-of-the-art prediction models
🔎 Similar Papers
No similar papers found.
