Center for Traffic Simulation Studies

This paper describes new adaptive fuzzy algorithms for coordinated ramp metering. The new model family named ACCEZZ (Adaptive and Coordinated Control of Entrance Ramps with Fuzzy Logic) was developed to overcome the limitations of existing coordinated ramp metering algorithms. Each model is explanined, evaluated via simulation, and compared to other ramp metering approaches in several scenarios. Coordinated ramp metering is achieved in the ACCEZZ models by applying fuzzy control to a series of entrance ramps where the interdependency of ramp operations is taken into account. A simple fuzzy ramp metering controller for each metered on-ramp is the core of each version of the ACCEZZ models. Learning/optimization methods drawn from both neural network theory and genetic algorithms are used to find the optimal ramp metering strategy. The resulting systems are either called neuro-fuzzy or genetic fuzzy ramp metering. The performance of the ACCEZZ models was assessed in a simulation context with a microscopic traffic flow model and compared with the results of five difference standard ramp metering algorithms: demand-capacity, occupancy strategy, ALINEA, Denver's HELPER algorithm and Minnesota's Zone approach. The total time spent in the system was used to evaluate the overall system performance of a strategy, since it includes both travel times and ramp delays. Additionally, the traffic densities, waiting times, queue lengths, fuel consumption and pollutants were compared. One of the ACCEZZ models will be installed in Munich, Germany, at the Olympic interchange of the ring road within the MOBINET project.