Optimal Coordination Methods for Autonomous Vehicles in Mixed Traffic

Doctoral thesis, 2025

Connected and Automated Vehicles (CAVs) are projected to dominate traffic roads in the future due to their potential advantages in efficiency and safety. CAVs are equipped with sensors and onboard computers that allows them to perform coordination. The transition toward fully autonomous era will see a gradual replacement of legacy Human-Driven Vehicles (HDVs) creating mixed traffic environments. In such environments, the presence of HDVs can pose challenges to CAVs due to their uncertain behaviors and intentions. The particular concern of CAVs-HDVs interactions occurs at traffic intersections, where these road segments are responsible for the highest share of traffic jams and fatalities. Additionally, vehicle coordination in mixed traffic involves computationally difficult problems that cannot be solved in a tractable way.

This thesis presents optimization-based coordination strategies which builds upon mixed-platooning scheme and heuristic approaches. By utilizing the CAVs presence, the platooning strategy is implemented to partially control the HDVs. To retrieve initial intersection crossing order, a feasibility-enforcing Alternating Direction Methods of Multipliers (ADMM) is employed. Furthermore, an optimization-based heuristic is developed to efficiently evaluate reordering scenarios. The heuristic employs constraint-feasibility check and cost comparison techniques. Next, in an economic optimal coordination scenario, a sensitivity-based heuristic is implemented to further reduce computational loads by approximating Nonlinear Program (NLP) solutions. The numerical results demonstrate that these heuristics can achieve near-optimal solutions and be better than the alternatives while can be hundred times faster than the Mixed-Integer Program (MIP) solvers.