1. Abstract

Autonomous driving systems face the formidable challenge of navigating intricate and dynamic environments with uncertainty. This study presents a unified prediction and planning framework that concurrently models short-term aleatoric uncertainty (SAU), long-term aleatoric uncertainty (LAU), and epistemic uncertainty (EU) to establish a robust foundation for planning in dynamic contexts. Using Gaussian mixture models and deep ensemble methods, our framework captures and assesses SAU, LAU, and EU simultaneously, surpassing traditional methods that treat these uncertainties separately. Additionally, we introduce an uncertainty-aware planning (UAP) approach that integrates these uncertainties into the decision-making process. Our contributions include comprehensive comparisons of uncertainty estimations, risk modeling, and planning methods against existing approaches. The proposed methods were rigorously evaluated using the CommonRoad benchmark and scenarios with limited perception, demonstrating significant improvements over existing approaches, particularly in diverse traffic scenarios. Comparative analyses highlight the advantages of incorporating multiple types of uncertainties to enhance planning accuracy and reliability. This study provides a detailed perspective on the application of uncertainty management from prediction to planning, and the findings underscore the potential for improved autonomous driving performance, especially in accident prevention, through comprehensive uncertainty management.

2. Method Overview

Proposed unified prediction and planning framework that considers different types of uncertainties.

The modeled uncertainties and their combinations, as well as various uncertainty-aware risk models.

3. Visaualization

4. Contact

If you have any questions, feel free to contact Wenbo Shao (swb19@mails.tsinghua.edu.cn).