Cooperative Optimal Control for Connected and Automated Vehicles in Mixed Traffic Merging Scenarios

Abstract

To enhance the safety of autonomous vehicles, research on connected autonomous vehicles (CAVs) that can communicate with surrounding vehicles and traffic infrastructure is actively being conducted. In mixed traffic environments, where actual CAVs will operate, controlling vehicles is challenging due to the difficulty in accurately predicting the future positions and speeds of human-driven vehicles (HDVs). To address this, machine learning-based predictive models using time-series data and decision trees have been utilized, but these methods face challenges in achieving scalability to all vehicle combinations.

This paper proposes an optimal control strategy with scalability for cooperative driving of CAVs in mixed traffic merging scenario. Cooperative driving is a technology that allows CAVs to drive safely and efficiently through V2X(vehicle-to-everything) communication. The proposed strategy focuses on determining the control actions of CAVs using a projection-based solution, allowing them to drive at speeds close to the target speed. The key idea is to solve an optimization problem that defines the movements of HDVs in terms of intersection arrival times, which calculates the intersection arrival times for CAVs through upper-level control, and using MPC(Model Predictive Control) for low-level control to manage the speeds of the CAVs. In particular, the upper-level control uses a projection-based solution to handle absolute value inequality constraints and to determine the intersection arrival times for vehicles merging from multiple roads into one. This approach simplifies the formulation and solution of the optimal control problem for complex scenarios, ensuring scalability.