Simulation Testing of a Fuzzy Neural Ramp Metering Algorithm

摘要

A fuzzy logic ramp metering algorithm will address the needs of Seattle's freewaysystem and overcome limitations of the existing ramp metering algorithm. The design of the fuzzy logic controller (FLC) reduced the sensitivity to sensor data which frequently contains errors or noise. The rule base effectively balanced two opposing needs: to alleviate mainline congestion by restriction the metering rate, and to disperse the ramp queue by increasing the metering rate. To avoid oscillation between these two conflicting demand, the controller used inputs that were more descriptive of congestion levels, providing smooth transitions rather than threshold activations. Testing was performed with freeway simulation software FRESIM. A multiple-ramp study site from Seattle's I-5 corridor was modeled using data such as freeway geometry entry volumes, desired speeds, and driver behavior. To evaluate the FLC under a variety of conditions, entry volumes and incidents (such as blocked lane or reduced capacity) were varied to create six test data sets. The performance of the FLC was compared to that of other available controller, including clock, demand/capacity, and speed metering. The objective was to maximize total vehicle miles, maximize mainline speeds, and minimize delay/vehicle-mile while maintaining an acceptable ramp queue. For five of the six data sets, the FLC outperformed the other three controllers. In the FLC, sensors from the on-ramp were helpful in maintaining an acceptable ramp queue. Future work will involve on-line testing of the FLC.