Dynamic High-Order Control Barrier Functions with Diffuser for Safety-Critical Trajectory Planning at Signal-Free Intersections

Di Chen,Ruiguo Zhong,Kehua Chen,Zhiwei Shang,Meixin Zhu,Edward Chung

2024-11-29

Abstract:Planning safe and efficient trajectories through signal-free intersections presents significant challenges for autonomous vehicles (AVs), particularly in dynamic, multi-task environments with unpredictable interactions and an increased possibility of conflicts. This study aims to address these challenges by developing a robust, adaptive framework to ensure safety in such complex scenarios. Existing approaches often struggle to provide reliable safety mechanisms in dynamic and learn multi-task behaviors from demonstrations in signal-free intersections. This study proposes a safety-critical planning method that integrates Dynamic High-Order Control Barrier Functions (DHOCBF) with a diffusion-based model, called Dynamic Safety-Critical Diffuser (DSC-Diffuser), offering a robust solution for adaptive, safe, and multi-task driving in signal-free intersections. Our approach incorporates a goal-oriented, task-guided diffusion model, enabling the model to learn multiple driving tasks simultaneously from real-world data. To further ensure driving safety in dynamic environments, the proposed DHOCBF framework dynamically adjusts to account for the movements of surrounding vehicles, offering enhanced adaptability compared to traditional control barrier functions. Validity evaluations of DHOCBF, conducted through numerical simulations, demonstrate its robustness in adapting to variations in obstacle velocities, sizes, uncertainties, and locations, effectively maintaining driving safety across a wide range of complex and uncertain scenarios. Performance evaluations across various scenes confirm that DSC-Diffuser provides realistic, stable, and generalizable policies, equipping it with the flexibility to adapt to diverse driving tasks.

Robotics,Machine Learning,Systems and Control

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to plan safe and efficient driving trajectories for autonomous vehicles at unsignalized intersections. Specifically, the research addresses the following challenges:

1. **Multi - task learning**: How to learn multiple driving tasks (such as left - turn, straight - ahead, right - turn) under different conditions.
2. **Safety - critical trajectory planning in complex scenarios**: How to conduct safety - critical trajectory planning in complex dynamic environments, especially in multi - task and highly interactive environments.

To solve these problems, the paper proposes a method that combines Dynamic High - Order Control Barrier Functions (DHOCBF) and diffusion models, called Dynamic Safety - critical Diffuser (DSC - Diffuser). The main contributions of this method include:

1. **Safety planning framework**: A framework integrating trajectory generation and safety correction modules is proposed for learning human driving strategies and ensuring safe driving at unsignalized intersections.
2. **Task - oriented and goal - oriented generation methods**: For the first time, task - oriented and goal - oriented generation methods are introduced into unsignalized intersections to recover strategies from interaction datasets.
3. **Dynamic High - Order Control Barrier Functions (DHOCBF)**: DHOCBF is proposed as a hard constraint to ensure that the ego - vehicle always stays within the safe area in a dynamic environment.
4. **Performance evaluation**: The performance of the proposed Dynamic Safety - critical Diffuser (DSC - Diffuser) algorithm is evaluated by comparing with existing algorithms and human driving trajectory data, and tested in different scenarios, demonstrating the strong generalization ability and safety of the model.

Through these methods and contributions, the paper aims to provide a robust and adaptable solution for the safe navigation of autonomous vehicles at unsignalized intersections.