Sustainable model-predictive control in urban traffic networks: Efficient solution based on general smoothening methods

Traffic-responsive control approaches, including model-predictive control (MPC), are efficient methods for making the best use of the available network capacity. Moreover, gradient-based approaches, which can be applied to smooth optimization problems, have proven their efficiency, both computationally and performance-wise, in finding optima of optimization problems. In this paper, we propose an MPC system for an urban traffic network that applies a gradient-based optimization approach to solve the control optimization problem. The
controller uses a new smooth integrated flow-emission model to find a balanced tradeoff between reduction of the congestion and of the total emissions. We also introduce efficient smoothening methods for nonsmooth mathematical models of physical systems.
The effectiveness of the proposed approach is demonstrated via a case study.