TAE: A Semi-supervised Controllable Behavior-aware Trajectory Generator and Predictor

TL;DR

This paper introduces TAE, a semi-supervised autoencoder that models driver behavior in a unified framework, enabling controllable trajectory generation and accurate prediction for safer autonomous vehicle planning.

Contribution

The novel behavior-aware TAE explicitly encodes driver behaviors in the latent space, unifying trajectory generation and prediction with controllability and enhanced scenario augmentation.

Findings

Achieves diverse and realistic trajectory generation.

Provides accurate behavior prediction alongside trajectories.

Enhances safety-critical scenario planning.

Abstract

Trajectory generation and prediction are two interwoven tasks that play important roles in planner evaluation and decision making for intelligent vehicles. Most existing methods focus on one of the two and are optimized to directly output the final generated/predicted trajectories, which only contain limited information for critical scenario augmentation and safe planning. In this work, we propose a novel behavior-aware Trajectory Autoencoder (TAE) that explicitly models drivers' behavior such as aggressiveness and intention in the latent space, using semi-supervised adversarial autoencoder and domain knowledge in transportation. Our model addresses trajectory generation and prediction in a unified architecture and benefits both tasks: the model can generate diverse, controllable and realistic trajectories to enhance planner optimization in safety-critical and long-tailed scenarios, and it can provide prediction of critical behavior in addition to the final trajectories for decision making. Experimental results demonstrate that our method achieves promising performance on both trajectory generation and prediction.