Sustainable and reliable design of autonomous driving lanes: A chance-constrained extended goal programming approach

As autonomous vehicles (AVs) advance and gradually integrate into urban roadways, managing traffic in mixed environments, where human-driven vehicles (HVs) coexist with AVs, has become a critical challenge. Previous research has developed several optimal network design models for autonomous driving lanes (ADLs) to partially address this issue. However, these studies largely overlook uncertainties in AV market share or travel demand, which may lead to either excessive investment in underutilized infrastructure or traffic inefficiencies with insufficient capacity. To address these challenges, this study introduces a chance constrained programming (CCP) approach with sample approximation to effectively incorporate such uncertainties while accommodating varying risk preferences. To harmonize various sustainable development goals, such as network efficiency and social equity, CCP is further combined with extended goal programming, forming the proposed chance-constrained extended goal programming (CCEGP) model in this study for sustainable and reliable ADL design. Additionally, the routing behaviors of HV and AV users, each with differing levels of traffic awareness, are unified within a mixed cross-nested logit-based stochastic user equilibrium model, thereby enhancing behavioral realism and model generalizability. The heuristic coati optimization algorithm is modified for solving, and case studies are conducted to validate its applicability.