Coordinated lane-changing scheduling of multilane CAV platoons in heterogeneous scenarios

Abstract

With the development of sensing, communication and automated driving technology, connected and automated vehicles (CAVs) are becoming promising solutions for future transport requirements. It is widely believed that a vehicle platoon is a good form to organize urban traffic in the CAV era. Due to the multicommodity nature of urban traffic streams, vehicles will continuously leave and join a multilane platoon, which inevitably gives rise to the need of lane changing within a multilane platoon. This paper studies the coordinated lane-changing scheduling problem in a CAV platoon, with the goal of transferring the platoon from an initial state to a target state to minimize a certain cost measurement (e.g., number of steps), while heterogeneous scenarios are considered. Two approaches, i.e., an exact and an approximate approach, are proposed. For the exact approach, we formulate an integer linear programming (ILP) model to identify the global optimal solution. Multiple objective functions are defined to meet the different needs. To relieve the computational issue of the exact approach, we further propose a tree-based heuristic search (THS), an approximate algorithm framework. THS is able to obtain an acceptable solution with negligible computational effort, and has the potential to handle the scheduling problem with more precise modeling or larger platoons. Numerical experiments are conducted to demonstrate the performance of different algorithms on both small- and large-scale cases (with up to 60 vehicles in a platoon), and the parameter combinations in the THS are tested for the optimal trade-off between solution quality and computational load. The findings indicate that ILP is practical for small- or medium-scale cases, which can generate multiple optimal solutions for different objectives; THS can solve large-scale cases in milliseconds on an ordinary personal computer, while the acquired solution is verified to be only slightly worse than the global optimum.