Towards Safe and Reliable Autonomous Driving: Dynamic Occupancy Set Prediction

TL;DR

The paper addresses the instability of trajectory predictions in autonomous driving by introducing Dynamic Occupancy Set (DOS) prediction that is tightly integrated with trajectory prediction networks. It advances with an elliptic DOS representation, a GRU-based DOS Prediction Module, and a multi-objective loss that balances sufficient coverage against compact DOS area, validated on CommonRoad and SIND where it outperforms PR-, SA-, and circle-based baselines. Key contributions include the DOS representation, the trajectory-prediction-aware DOS framework, and new metrics (CR and OSA) showing improved safety and planning efficiency. The work has practical significance for reliable autonomous navigation in dynamic, uncertain environments and offers a clear path for integrating DOS into planning stacks.

Abstract

In the rapidly evolving field of autonomous driving, reliable prediction is pivotal for vehicular safety. However, trajectory predictions often deviate from actual paths, particularly in complex and challenging environments, leading to significant errors. To address this issue, our study introduces a novel method for Dynamic Occupancy Set (DOS) prediction, it effectively combines advanced trajectory prediction networks with a DOS prediction module, overcoming the shortcomings of existing models. It provides a comprehensive and adaptable framework for predicting the potential occupancy sets of traffic participants. The innovative contributions of this study include the development of a novel DOS prediction model specifically tailored for navigating complex scenarios, the introduction of precise DOS mathematical representations, and the formulation of optimized loss functions that collectively advance the safety and efficiency of autonomous systems. Through rigorous validation, our method demonstrates marked improvements over traditional models, establishing a new benchmark for safety and operational efficiency in intelligent transportation systems.

Figures (5)

• Figure 1: DOS prediction and its design objectives.
• Figure 2: Trajectory-prediction-aware DOS prediction model.
• Figure 3: Design of DOS prediction module.
• Figure 4: Analysis Different DOS Representations.
• Figure 5: Visualization of Different DOS Representations. Red or green Regions Represent Predicted DOS at Each Time.