TAE: A Semi-supervised Controllable Behavior-aware Trajectory Generator and Predictor
This work addresses the problem of limited information in trajectory generation and prediction for intelligent vehicle planning, offering a unified approach to enhance safety-critical and long-tailed scenarios, though it is incremental as it builds on existing autoencoder and adversarial methods.
The paper tackles the intertwined tasks of trajectory generation and prediction for intelligent vehicles by proposing a behavior-aware Trajectory Autoencoder (TAE) that models driver behavior like aggressiveness and intention in a latent space, achieving promising performance in both tasks.
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.