Semi-autonomous Intersection Collision Avoidance through Job-shop Scheduling

Abstract: In this paper, we design a supervisor to prevent vehicle collisions at intersections. An intersection is modeled as an area containing multiple conflict points where vehicle paths cross in the future. At every time step, the supervisor determines whether there will be more than one vehicle in the vicinity of a conflict point at the same time. If there is, then an impending collision is detected, and the supervisor overrides the drivers to avoid collision. A major challenge in the design of a supervisor as opposed to an autonomous vehicle controller is to verify whether future collisions will occur based on the current drivers choices. This verification problem is particularly hard due to the large number of vehicles often involved in intersection collision, to the multitude of conflict points, and to the vehicles dynamics. In order to solve the verification problem, we translate the problem to a job-shop scheduling problem that yields equivalent answers. The job-shop scheduling problem can, in turn, be transformed into a mixed-integer linear program when the vehicle dynamics are first-order dynamics, and can thus be solved by using a commercial solver.

• The paper presents a novel method that transforms intersection collision avoidance into a MILP job-shop scheduling problem to ensure safety.
• It demonstrates real-time performance by computing solutions within 100 ms per iteration even in complex urban scenarios.
• Simulation results confirm the supervisor system effectively overrides driver inputs while managing up to twenty vehicles per intersection.

Semi-Autonomous Intersection Collision Avoidance through Job-Shop Scheduling

Introduction

The paper "Semi-autonomous Intersection Collision Avoidance through Job-shop Scheduling" presents a novel approach to prevent vehicle collisions at intersections by employing a semi-autonomous supervisor system. The authors address the challenges inherent in intersection collision avoidance due to the complex vehicle dynamics and the multitude of conflict points where vehicle paths intersect. Traditional autonomous vehicle controllers face difficulties verifying future collisions based on current driver choices, necessitating a system that can minimally override driver inputs only when a collision is imminent.

Methodology

The authors propose transforming the problem of intersection collision avoidance into a job-shop scheduling problem, which can be further converted into a mixed-integer linear programming (MILP) problem. This transformation allows for the real-time execution of collision avoidance strategies using commercial solvers. In essence, the supervisor system functions by detecting impending collisions through analysis of vehicle trajectories and conflict point dynamics, overriding driver controls selectively to prevent these collisions.

The supervisor relies on a model of the intersection, comprising various conflict areas, each defined by potential points of vehicle path intersection. By translating the vehicle dynamics and driver inputs into scheduling variables, the authors harness the structure of job-shop scheduling to navigate computational complexity. Despite being an NP-hard problem, the scheduling approach is feasible due to the conversion into MILP form, which finds efficient solutions using solvers such as CPLEX and Gurobi.

Results

Simulation results demonstrate the supervisor’s capability to maintain safety in complex intersection scenarios, such as those represented by Massachusetts' twenty most dangerous intersections. The supervisor effectively overrides dangerous driver actions, ensuring collision-free paths through the intersection. Notably, computation times for the algorithm remain within the practical limit of 100 ms per iteration, underscoring the feasibility and efficiency of the approach. The data indicate that the supervisor can handle typical intersections with up to twenty vehicles, a scenario common in real-world urban environments.

Implications and Future Work

This research presents significant implications for semi-autonomous vehicle systems in urban environments. The supervisor provides a safety net for intersections without requiring full autonomy, thus integrating smoothly with existing traffic systems. The approach can serve as a transitional technology toward fully autonomous vehicles by enhancing current infrastructure with collision avoidance capabilities.

Future work will explore extensions to include rear-end collision considerations and second-order vehicle dynamics to improve realism. Additionally, addressing measurement uncertainties and unequipped vehicle scenarios will enhance robustness. The possibility of accommodating undetermined routes and steering input decisions further expands the potential applications of this scheduling approach.

Conclusion

"Semi-autonomous Intersection Collision Avoidance through Job-shop Scheduling" offers a compelling intersection management solution capable of ensuring safety with minimal disruption to driver autonomy. The transformation of collision avoidance into a solvable scheduling problem represents a practical advance, paving the way for improved traffic safety systems in urban settings. Through computational efficiency and strategic override capabilities, this research bridges the gap between current vehicle dynamics and the future of intelligent infrastructure solutions.