System-Level Evaluation of Autonomous Vehicle Lane Deployment Strategies Under Mixed Traffic Flow

Abstract

Connected and Autonomous Vehicles (CAVs) are expected to reshape future transportation systems. During the long transition period, in which CAVs and human-driven vehicles (HVs) coexist, deploying CAV-dedicated lanes offers a promising approach to enhancing overall efficiency, but raises concerns about distributional fairness. This study develops a system-level evaluation framework that integrates bi-level network capacity optimization with practical planning constraints to determine optimal lane-deployment strategies. The bi-level model aims to maximize network reserve capacity at the upper level, while it captures mixed-traffic flow distribution under the lower-level user equilibrium (UE) principle. Both levels are constrained by CAV market penetration (MPR), social equity, and budget bound considerations. To ensure computational tractability, nonlinear relationships are linearized through Piecewise Linear Approximation (PLA), converting the original Mixed-Integer Nonlinear Programming (MINLP) model into a Mixed-Integer Linear Programming (MILP) formulation solvable by standard optimization solvers. Numerical experiments on the Sioux Falls network demonstrate that increasing MPR and dedicated lane deployment can substantially improve network capacity by up to 36% compared with the baseline, with diminishing marginal benefits as deployment scale excesses. Incorporating equity constraints further reduce the HV–CAV cost gap, promoting fairer outcomes without significant efficiency loss. These findings offer quantitative evidence on the efficiency–equity trade-offs in CAV-dedicated lanes planning and provide practical implications for policymakers in developing sustainable strategies.