A Computationally Efficient Bi-level Coordination Framework for CAVs at Unsignalized Intersections

TL;DR

This paper introduces a bi-level coordination framework for connected and automated vehicles at unsignalized intersections, combining high-level priority scheduling with low-level trajectory planning to improve traffic throughput and computational efficiency.

Contribution

It proposes a novel bi-level coordination approach using Monte Carlo Tree Search and priority-based trajectory planning, enhancing scalability and performance over existing methods.

Findings

Achieves near-optimal traffic coordination performance.

Significantly outperforms traditional traffic signal control systems.

Scales linearly with the number of vehicles.

Abstract

In this paper, we investigate cooperative vehicle coordination for connected and automated vehicles (CAVs) at unsignalized intersections. To support high traffic throughput while reducing computational complexity, we present a novel collision region model and decompose the optimal coordination problem into two sub-problems: \textit{centralized} priority scheduling and \textit{distributed} trajectory planning. Then, we propose a bi-level coordination framework which includes: (i) a Monte Carlo Tree Search (MCTS)-based high-level priority scheduler aims to find high-quality passing orders to maximize traffic throughput, and (ii) a priority-based low-level trajectory planner that generates optimal collision-free control inputs. Simulation results demonstrate that our bi-level strategy achieves near-optimal coordination performance, comparable to state-of-the-art centralized strategies, and significantly outperform the traffic signal control systems in terms of traffic throughput. Moreover, our approach exhibits good scalability, with computational complexity scaling linearly with the number of vehicles. Video demonstrations can be found online at \url{https://youtu.be/WYAKFMNnQfs}.