Integrated optimal planning of multi-type lanes and intersections in a transportation network with mixed HVs and CAVs

Abstract

In a prolonged transitional period, the urban transportation infrastructure is expected to accommodate two different types of vehicles. One type of vehicle is the human-driven vehicle (HV) and the other is the connected and autonomous vehicle (CAV). Nevertheless, the conflict between HVs and CAVs in the mixed traffic scenario significantly impedes the efficiency-improvement benefit of implementing emerging CAV technologies. To harness the full potential of CAVs in enhancing traffic efficiency and network performance, multi-type lanes (i.e., regular lanes, dedicated CAV lanes, and CAV/toll lanes) and multi-type intersections (i.e., conventional signalized intersections, novel signalized intersections with an exclusive phase and exclusive approaches, and smart signal-free intersections) are proposed to efficiently manage the heterogeneous traffic flow on roads and at intersections, respectively. From the perspective of traffic planners, in this research, the integrated planning problem of multi-type lanes and intersections (IPPLI for short) in the mixed transportation network is suggested and tackled, where the route selection behavior and the cross-group externalities of heterogeneous travelers are considered according to the user equilibrium principle. It aims to minimize the overall travel cost by making decisions on the spatial layout of multi-type lanes and intersections in the network, the toll level of CAV/toll lanes, the number of exclusive approaches at novel signalized intersections, time intervals of the cycle and green signal for each phase at both conventional and novel signalized intersections. Then, the IPPLI is formulated as a mixed-integer nonlinear programming model based on the link-node modeling method without time-consuming path enumeration and memory-consuming path storage. As a mathematical problem with complementarity constraints, it is solved by an improved evolutionary algorithm-based approach, which consists of two modules cooperating with each other. After introducing the concept of the accessibility of HVs, a heuristic technique is proposed to accelerate algorithm convergence by continuously repairing unreasonable solutions. Finally, experiments are performed on two distinct networks to showcase the properties of the problem and assess the effectiveness of the proposed model. Experimental results show that the proposed model consistently performs outstandingly across a range of CAV penetration rates. Our model achieves maximum improvements of 25.71% and 4.84% in reducing travel costs compared to models that only plan multi-type lanes and multi-type intersections, respectively. Additionally, the improved evolutionary algorithm-based approach reduces the convergence time by 20.51% and 26.81% compared to the classical evolutionary algorithm and the genetic algorithm provided by MATLAB Global Optimization Toolbox.