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

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 behavioraware 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.

Semi-supervised Semantics-guided Adversarial Training for Trajectory Prediction

“…The encoder maps high-dimensional features to the semantics-guided latent space with distribution regularization and the semi-supervised training. In particular, the latent space is divided into three parts: longitudinal features z lon , following one-dimensional log-normal distribution, lateral features z lat , following three-dimensional categorical distribution and remaining features, following Gaussian distribution [15]. Finally, the decoder map semantic vectors along with other disentangled latent vectors to the future trajectories.…”

“…The trajectory prediction plays a crucial role for understanding the environment and making safety-critical decisions in the following planning module. Recent works [11,12,13,14,15] have applied various deep learning techniques to this task and achieved impressive performance in terms of reducing average errors. Some works [11,12,13,14] utilize advanced feature extractors such as graph neural networks or transformers, and the work in [15] proposes a behavior-aware trajectory generator.…”

“…Recent works [11,12,13,14,15] have applied various deep learning techniques to this task and achieved impressive performance in terms of reducing average errors. Some works [11,12,13,14] utilize advanced feature extractors such as graph neural networks or transformers, and the work in [15] proposes a behavior-aware trajectory generator. However, few works studied the robustness of vehicle trajectory prediction, which is in fact of vital importance because 1) autonomous driving is a safety-critical task by nature, 2) recent work [16] demonstrates that the prediction module is vulnerable to adversarial attacks if surrounding vehicles drive along a seemingly natural but crafted trajectory, and 3) current prediction models are often overfitted on the limited patterns in the datasets but suffer the long-tailed distribution of driving scenarios and behaviors.…”

Efficient Speed Planning for Autonomous Driving in Dynamic Environment With Interaction Point Model

“…For the discussed POMDPs [3], [4], [10], interactions were reflected by time-consuming simulations but not modeled directly. In recent years, learning-based approaches [19], [20] have been promising to learn trajectory generation and prediction in a unified architecture. However, machine learning technologies lacked traceability when an error occurred in practice.…”

Learning Representation for Anomaly Detection of Vehicle Trajectories

“…In particular, the prediction module encodes other vehicles' past trajectories along with map context and decodes them into potential future trajectories of surrounding vehicles to facilitate the planning module. Recent works [1]- [4] have developed various deep learning-based models for trajectory prediction and achieved great performance in terms of the average error between predicted trajectories and ground truth. However, only improving average performance is not enough for autonomous driving systems, where system robustness, safety and security are critical.…”