Cloud-based Distributed Data-Enabled Predictive Control for Time-Varying Mixed Traffic Flow

This paper addresses the optimal control problem for mixed traffic flow systems with unknown dynamics and diverse system constraints. We propose a cloud-based distributed Online Data-Enabled Predictive Control (ODeePC) algorithm to regulate traffic flow toward equilibrium states. To overcome the limitations of traditional data-driven methods in handling time-varying system dynamics, real-time data samples are incorporated to update the problem representation while retaining a subset of offline samples. By combining real-time adaptation with historical data, the proposed algorithm enhances the robustness of control while preserving the feasibility of the solution. To efficiently handle the large-scale problem, we employ the Alternating Direction Method of Multipliers (ADMM) to decompose the optimization task into smaller subproblems, which are solved via iterative computation. Minimal information exchange between subsystems occurs to achieve global optimization. Leveraging the extensive computational power and parallel processing capabilities of the cloud, the proposed algorithm is implemented and deployed as a workflow-based cloud service. Dedicated communication-resilience strategies are also designed to address the delays and packet loss inherent in cloud deployment. The effectiveness of the proposed framework is demonstrated through comprehensive evaluations based on simulation comparisons and real-world deployment in a cloud environment. The results highlight the potential of cloud-based, distributed control for real-time management of complex traffic systems.