Novel and efficient local coordinated freeway ramp metering strategy with simultaneous perturbation stochastic approximation-based parameter learning

In this study, a novel and efficient fuzzy logic control (FLC) -based local coordinated ramp metering strategy is studied for networked freeway traffic control. Local traffic measurements are exchanged among neighbouring controllers. Based on both local and exchanged traffic measurement, the FLC based coordinated decision-making algorithms determine the reference mainstream densities, which are consequently tracked by local controllers using feedback control-based strategy. The parameters of the FLC algorithm are adjusted by a simple and efficient simultaneous perturbation stochastic approximation based parameter learning algorithm, which finds the optimal coordinated decision-making policies without interruption to normal freeway operations. Due to the limited communication among neighbouring controllers, communication cost is reduced compared with existing centralised control-based strategies. Furthermore, the structure of the proposed strategy is simple without involvement of complex model-based calculation. Case studies show the potential of the proposed strategy for coordinated freeway ramp metering problems with significantly improved system efficiency. In particular, using the proposed local coordinated ramp metering strategy, the system cost function is reduced by over 48% compared with no control case, and by over 18% compared with uncoordinated feedback control-based ramp metering case. The queue volumes at the on-ramp links are well balanced both temporally and spatially.