Munhoz Arfvidsson, Kaj (KTH Royal Institute of Technology), Jiang, Frank J. (KTH Royal Institute of Technology), Johansson, Karl H. (KTH Royal Institute of Technology), Mårtensson, Jonas (KTH Royal Institute of Technology)

Towards Safe Autonomous Intersection Management: Temporal Logic-Based Safety Filters for Vehicle Coordination

Scheduled for presentation during the Invited Session "Traffic Control and Connected Autonomous Vehicles: benefits for efficiency, safety and beyond (2 edition) II" (WeBT4), Wednesday, September 25, 2024, 15:50−16:10, Salon 8

2024 IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), September 24- 27, 2024, Edmonton, Canada

This information is tentative and subject to change. Compiled on October 3, 2024

Abstract

In this paper, we introduce a temporal logic-based safety filter for Autonomous Intersection Management (AIM), an emerging infrastructure technology for connected vehicles to coordinate traffic flow through intersections. Despite substantial work on AIM systems, the balance between intersection safety and efficiency persists as a significant challenge. Building on recent developments in formal methods that now have become computationally feasible for AIM applications, we introduce an approach that starts with a temporal logic specification for the intersection and then uses reachability analysis to compute safe time-state corridors for the connected vehicles that pass through the intersection. By analyzing these corridors, in contrast to single trajectories, we can make explicit design decisions regarding safety-efficiency trade-offs while taking each vehicle's decision uncertainty into account. Additionally, we compute safe driving limits to ensure that vehicles remain within their designated safe corridors. Combining these elements, we develop a service that provides safety filters for AIM coordination of connected vehicles. We evaluate the practical feasibility of our safety framework using a simulated 4-way intersection, showing that our approach performs in real-time for multiple scenarios.