LOCALIZED SWITCHING RAMP-METERING CONTROL WITH QUEUE LENGTH ESTIMATION AND REGULATION AND MICROSCOPIC SIMULATION RESULTS ⋆ Xiaotian Sun 1 Roberto Horowitz 2 Department of Mechanical Engineering University of California at Berkeley Berkeley, CA 94720-1740, USA Abstract: In this paper, we firstreview a localized ramp-metering strategyfor freeways that achieves the goal ofreducing the spatial and temporal span of traffic congestion, using locally available information.The control scheme is composed of a switching mainline-traffic responsive ramp-metering controllerthat adapts to the different traffic dynamics under different congestion conditions—free-flow or congested—and a queue length regulator that yields improved performance over the currently used ad hoc “queue- override” scheme and prevents the queue from exceeding the ramp storage capacity. Subsequently, a queue length estimatoris designed toprovide feedback to the queue length regulator, using the queue-detector speed data that areavailable in the field.T est results on a calibrated microscopic trafficsimulator demonstrate the performance and effectiveness ofthe switching ramp-metering controller, the queue length estimator and regulator, and the overall control strategy. The T otalV ehicle and Passenger Delaysare both reduced by16%, while the T otalV ehicle Time and the T otal A verage V ehicle Speed are improved by 5.6% and 5.8%, respectively. As a comparison, simulation resultsof ALINEA are also presented. Copyright © 2005 IFAC Keywords: Transportation Systems, Traffic Control, On-Ramp Metering, Switching Control, Queue Length Estimation, Microscopic Simulation 1. INTRODUCTION Freeway traffic congestion is a major problem in today’ s metropolitan areas. It occurs regularly during commute hours. In addition, non-recurrent congestion often takes place as a result ofincidents,road work, or public events. Congestion causes inefficient operation of freeways, wasting ofresources,increased pollution, and intensified driver fatigue. The 2004 Urban Mobility Report (Schrank and Lomax, 2004) finds:“Congestion has grown everywhere in areas of all sizes. Congestion occurs during longer portions ofthe day and delays more travelers and goods ⋆ This work has been supported by California Partners for Advanced Transit and Highways (PATH) under Task Orders 4136 and 5503. 1 E-mail: sunx@me.berkeley.edu. 2 Corresponding author. E-mail: horowitz@me.berkeley.edu, tel: +1-510-642-4675, fax: +1-510-643-5599. than ever before.” In the report, it was calculated that in2002, congestion cost Americans 3.5 billion hours of delay and 5.7 billion gallons of wasted fuel, with an equivalent monetary cost of U.S. $63.2 billion. On-ramp metering has been widely used as an effective strategyto increase freeway operation efficiency. It has been recommended to the U.S. Federal Highway Administration as the No.1toolto address the conges- tion problem, otherthan adding more capacity to trans- portation infrastructures (Cambridge Systematics, Inc. and Texas Transportation Institute, 2004). It has been reported thatramp metering was able to reduce delay by 101 million person-hours in2002, approximately 5% ofthe congestion delay on freeways where ramp- metering was in effect (Schrank and Lomax, 2004). In this paper, we first briefly review a set oflocalized ramp-metering algorithms that we have developed (Sun and Horowitz, 2005), including a switching mainline-