On Distributed Optimal Control of Traffic Flows in Transportation Networks

We propose and analyze distributed computation algorithms for finite-horizon optimal control problems in transportation networks. We model traffic flow dynamics by the cell-transmission model and focus on two problems: system-optimum dynamic traffic assignment (where the routing is part of the optimization) and freeway network control (where the routing is exogenous and the optimization is confined to speed limits and ramp-metering controls). While these are non-convex problems, we focus on some recently proposed provably exact convex relaxations and apply Alternating Direction Method of Multipliers techniques. We present fully distributed iterative algorithms and implement them on some transportation network testbeds, testing their convergence speed and accuracy