Abstract—Freeway corridor traffic flow is limited by bottleneck flow. A possible approach for maximizing recurrent bottleneck flow is to create a discharge section immediately upstream of the bottleneck. This paper proposes a control strategy for combining Variable Speed Limits (VSL) and Coordinated Ramp Metering (CRM) design to achieve this objective when the bottleneck can be modeled as a lane drop (or virtual lane drop) and/or weaving section. The control design strategy is to design VSL first and then use Finite Time Horizon Model Predictive Control to design Coordinated Ramp Metering. Microscopic simulation results showed that the control strategy could improve traffic throughput significantly. I. INTRODUCTION Ramp metering is the most widely practiced strategy to control freeway traffic in the US, particularly in California. It is recognized that ramp metering can directly control the flow into the freeway and the average density immediately downstream of the onramp. After entering the freeway, the collective behaviors of the drivers are not controlled. This is why using ramp metering alone to control freeway traffic has limited performance if the flows from onramps and the upstream mainline are high. In addition, from the perspective of equity among the onramps along a corridor and the ramp queue length limit due to road geometry, ramp metering has to be switched off if the demand from that onramp is too high to avoid traffic spilling back onto arterials. Therefore, from a systems and control viewpoint, using ramp metering alone cannot fully control the freeway traffic in practice. This is the motivation for investigating other control strategies such as Variable Speed Limits. VSL attempts to control the collective vehicle speed (or driver behavior) of mainline traffic, which is complementary to RM. This work was supported by the Federal Highway Administration (FHWA) Exploratory Advanced Research Program (Cooperative Agreement DTFH61-07-H-00038) with matching funding from the California Department of Transportation (Caltrans). X. Y. Lu (Corresponding Author), is with PATH, ITS, U. C. Berkeley, Richmond Field Station, Bldg 452, 1357 S. 46th Street, Richmond, CA 94804, Tel: 1-510-665 3644, (xylu@path.berkeley.edu) P. Varaiya is with Dept of EECS, U. C. Berkeley, 271M Cory Hall, Berkeley CA 94720, Tel: 1-510-642-5270, (variya@eecs.berkeley.edu) R. Horowitz is with Dept of ME, University of California, Berkeley 5138 Etcheverry Hall, Berkeley CA 94720, Tel: 1-510- 642-4675, (horowitz@me.berkeley.edu) Dongyan Su is a GSR in ME, U. C. Berkeley, (dongyan@berkeley.edu) S. Shladover is with PATH, U. C. Berkeley, Tel: 1-510-665 3514, (steve@path.berkeley.edu). The following acronyms are used throughout the paper: VSL – Variable Speed Limit; RM – Ramp Metering; CRM – Coordinated RM; CTM – Cell Transmission Model; FD – Fundamental Diagram; TOPL (Tools for Operational Planning); SWARM - System Wide Adaptive Ramp Metering ; TTT – Total Travel Time; TTS - Total Time Spent; TTD – Total Traveled Distance; MPC – Model Predictive Control; EB (WB) – East Bound (West Bound); Several implementations have been conducted in the UK, France, Germany and Netherlands using VSL to harmonize the traffic mainly for safety rather than for mobility improvement. It is generally accepted that crashes and incidents can be reduced between 25~40 percent [1]. However, none of the VSL practice reported before was intended for maximizing traffic flow. This paper will focus on mobility improvements along a stretch of freeway using combined VSL and RM. Freeway traffic flow is limited by bottleneck flow. The causes of bottleneck may vary from case to case. In general, bottlenecks can be classified as: recurrent - the location and congestion time are predictable; non-recurrent– location and time are non-predictable. To maximize freeway traffic flow, one possible approach is to maximize the bottleneck flow. The main strategy to achieve this is to create a discharging section right before the bottleneck such that the feeding flow into the bottleneck could be closer its capacity flow. This consideration is based on previous works [2-7], which indicate that congested upstream traffic will reduce the bottleneck flow by about 5~20% depending on location and times. There are several possible ways to combine VSL and RM depending on what model is adopted and how the control strategy is designed, classified as follows:  Determine RM before determining VSL;  Determine RM and VSL simultaneously with tightly coupled speed and density dynamics model;  Determine VSL first before determining RM rate. In the work of [8], RM was designed before VSL or assumed known. It has some practical implication since RM has been widely implemented in many states in the U.S., particularly in California. This paper uses the third approach to design a combined traffic control strategy for maximizing the recurrent bottleneck flow. It determines VSL for each link/cell without a model while taking into account the following factors: maximizing the bottleneck flow, mainline and onramp demand variation, onramp length limit (storage A New Approach for Combined Freeway Variable Speed Limits and Coordinated Ramp Metering Xiao-Yun Lu, Pravin Varaiya (Fellow IEEE), Roberto Horowitz, Dongyan Su, and Steven E. Shladover