Resilient Transportation: An Integrated Corridor Management Approach

Transcription

1 Gulf Coast Research Center for Evacuation and Transportation Resiliency LSU / UNO University Transportation Center Resilient Transportation: An Integrated Corridor Management Approach Final Report Sherif Ishak, Ph.D. Syndney Jenkins, Undergraduate Student Danhong Cheng, Graduate Student Julius Codjoe, Graduate Student Performing Organization Gulf Coast Research Center for Evacuation and Transportation Resiliency Merritt C. Becker Jr. University of New Orleans Transportation Institute New Orleans, LA Sponsoring Agency Department of Transportation Research and Innovative Technology Administration Washington, DC Project # May 2011

2

3 Technical Report Documentation Page 1. Report No. 2. Government Accession No. 3. Recipient s Catalog No. 4. Title and Subtitle Resilient Transportation: An Integrated Corridor Management Approach 7. Author(s) Sherif Ishak, Ph.D., Syndney Jenkins, Danhong Cheng, and Julius Codjoe 9. Performing Organization Name and Address Department of Civil and Environmental Engineering, Louisiana State University Baton Rouge, LA Sponsoring Agency Name and Address Gulf Coast Center for Evacuation and Transportation Resiliency (GCCETR) Department of Civil and Environmental Engineering Louisiana State University Baton Rouge, LA Report Date 6. Performing Organization Code 8. Performing Organization Report No. 10. Work Unit No. (TRAIS) 11. Contract or Grant No. 13. Type of Report and Period Covered 14. Sponsoring Agency Code 15. Supplementary Notes 16. Abstract The primary goal of this research is to lay a foundation for the application and implementation of integrated corridor management (ICM) strategies to reduce congestion on the freeway and arterial systems in Baton Rouge. Ramp metering was identified as one of the successful strategies for ICM, and therefore, was selected for evaluation in this study. A dual freeway corridor, encompassing I-10 and I-12 in the area of Baton Rouge, was selected as a study area where ramp meters were installed at every on-ramp in both directions. A microscopic simulation model, VISSIM, was used to model traffic behavior under two scenarios: one with activated ramp meters and one without ramp meters. Performance measures were selected and generated from 25 randomly generated runs for each simulation scenario. Comparative evaluation was conducted to determine the benefits of applying ramp metering strategies. The results strongly suggest that the overall network performance has substantially improved with the implementation of ramp meters. Based on the selected performance measures, the comparative evaluation of both scenarios (with and without ramp metering) shows a statistically significant improvement in the corridor performance when ramp metering strategies are implemented. The statistical analysis using the Student s t-test for two independent samples with unknown variances showed consistently that the means were significantly different at 95% confidence level. A test of variances was also conducted and concluded that both populations had equal variances, and therefore, a pooled t-test analysis was conducted. Also, Fisher s Least Significant Difference (LSD) method was applied on the difference between the two population means at 95% confidence level. Based on the simulation results, the study recommends the use of ramp metering on both segments of I-10 and I Key Words Ramp metering, traffic simulation, freeway operation. 18. Distribution Statement No restrictions. Copies available from GCCETR: Security Classification (of this report) Unclassified 20. Security Classification (of this page) Unclassified 21. No. of Pages Price i

4

5 ACK OWLEDGEME TS This project was funded by the Gulf Coast Center for Evacuation and Transportation Resiliency (GCCETR). The authors wish to thank all who contributed to the completion of the project, including graduate students. Special thanks go to Thomas Montz for his help in getting the forecasted traffic data using TRA SCAD. The authors are also indebted to Mr. Huey Dugas, the director of Baton Rouge Capital Region Planning Commission for providing the travel demand forecasts for the I-10 and I-12 corridors in Baton Rouge. DISCLAIMER The contents of this report reflect the views of the authors, who are solely responsible for the facts and the accuracy of the material and information presented herein. This document is disseminated under the sponsorship of the U.S. Department of Transportation University Transportation Centers Program in the interest of information exchange. The U.S. Government assumes no liability for the contents or use thereof. The contents do not necessarily reflect the official views of the U.S. Government. This report does not constitute a standard, specification, or regulation. 3

6 1. BACKGROU D A D OBJECTIVES PROBLEM STATEME T Traffic congestion continues to escalate and pervade surface transportation systems in the US. Symptoms can be observed in a number of large and medium size cities across the country as travel demand continues to exceed the existing network capacity. Conventional approaches relying primarily on capacity expansion are high cost solutions that repeatedly failed to cope with the continuously rising demand and limited rights of way needed for roadway widening. Lately, transportation professionals recognized the need for better management of the existing network capacity as a viable alternative to capacity expansion investments. Transportation corridors may still have unused capacity on parallel routes that can be leveraged to alleviate congestion on freeways. Such concept has been referred to as Integrated Corridor Management (ICM) and successfully applied to major metropolitan areas such as Dallas, Texas, Houston, Texas, Minneapolis, Minnesota, Oakland, California, Seattle, Washington, to name a few. The city of Baton Rouge, the state capital of Louisiana, continues to grow in population and in travel demand at an alarming rate. The existing capacity of the roadway infrastructure in the city cannot sustain the rising demand, and therefore, congestion can now be seen over the freeway segments of I-10 and I-12, as well as the main arterials. Such congestion may be alleviated by applying integrated corridor management strategies such as ramp metering, which is the focus of this proposed research. RESEARCH OBJECTIVES The primary goal of this research is to lay the foundation for the application and implementation of integrated corridor management strategies to reduce congestion on the freeway and arterial systems in Baton Rouge. Under the ICM umbrella, the operation of freeways and arterials should be optimized for various functions such as traffic incident management, work zone management, planned special events management, and recurrent day-to-day conditions. This goal ensures more sustainable and resilient transportation network under both normal and extreme (such as emergency evacuation) operating conditions. It is possible to develop an efficient integrated corridor management by developing a ramp-metering strategy, information dissemination strategy and other ITS strategies along congested corridors. Due to the exponential increase in population and travel demand, Baton Rouge may potentially benefit from integrated corridor management strategies such as ramp metering which essentially targets congestion mitigation. This study applies ramp metering strategies on the two corridors of I-10 and I-12 within the city of Baton Rouge in order to determine their effectiveness. This is achieved by simulating both corridors with and without ramp metering at the microscopic level using the forecasted traffic demand in the year The specific objectives of this phase of research are to: Review the state of the practice of ramp metering strategies and their application in other metropolitan areas in order to learn from similar experiences and identify the various strategies used thus far, as well as their points of strengths and weaknesses. 4

7 Identify the data requirements for developing a simulation model for the two corridors of I-10 and I-12 in Baton Rouge. Forecasted travel demand is used for the year Select a microscopic simulation platform (e.g. VISSIM/DYNASMART) and build the simulation network for the study area. Evaluate the effectiveness of ramp metering by comparing selected network performance measures with and without the implementation of ramp meters. SCOPE The scope of the study is limited to the city of Baton Rouge in Louisiana. The study area boundary includes the main freeway system (I-10 and I-12). BACKGROU D This section presents a preliminary review of some of the research studies on ramp metering. Ramp metering aims to improve the traffic conditions by regulating the inflow from the onramps to the freeway main stream. For fixed time metering strategies, ramp meter timings are adjusted for different time periods during the day, and therefore, do not offer the flexibility to adapt to changing traffic conditions. Traffic-responsive ramp metering strategies, on the other hand, are based on real-time measurements from sensors installed in the freeway network and can be classified as local or coordinated. Local control is a process of selecting ramp metering rates based solely on conditions present at an individual ramp, while coordinated control is a process of selecting metering rates based on conditions throughout the entire metered corridor. Local Ramp Metering Strategies Masher et al. (1975) developed a demand-capacity ramp metering algorithm, which is a traffic responsive algorithm that measures the downstream occupancy. If the occupancy exceeds a critical value, congestion is assumed to exist and the metering rate is then set to the min rate. Otherwise, the volume is measured upstream of the merge, and the metering rate is set to the difference between the downstream capacity and the upstream volume. Papageorgiou et al. (1997) proposed a local responsive feedback ramp metering strategy (ALINEA), which had multiple successful field applications (Paris, Amsterdam, Glasgow, Munich). This algorithm considers traffic flow as the process being controlled and the metering rate as the control variable. Based on feedback control theory, the algorithm attempts to set the metering rate such that traffic flow will not exceed system capacity. The algorithm uses the difference in occupancy values (desired or capacity versus measured), measured at a point 40 meters downstream of the ramp gore, to calculate a metering rate. One of the most desirable features of this algorithm is the integration of the previous time interval metering rate within the equation. This allows integrated smoothing of the metering rates to avoid wide swings between concurrent time intervals. In another paper, Smaragdis et al. (2003) presented several modifications and extensions of ALINEA. Specifically, FL-ALINEA is a flow-based strategy; UP-ALINEA is an upstream occupancy - based version; UF-ALINEA is an upstream-flow-based strategy. X-ALINEA/Q is 5

8 the combination of any of the above strategies with efficient ramp-queue control to avoid interference with surface street traffic. A zone algorithm was reported to be used at Minnesota (Thompson et al, 1997). This algorithm defines directional freeway facility metering zones with zones having variable lengths of three to six miles. Its basic concept of the algorithm is to balance the volume of traffic entering and leaving each zone. All entering and exiting traffic volumes on both the mainline and the ramps are measured in 30-second increments, and balancing these total volumes is used to keep the density of traffic within the zone constant. Ghods et al. (2009) proposed an adaptive genetic fuzzy control approach to reduce peak hour congestion, along with speed limit control. To calibrate the fuzzy controller, a genetic algorithm is used to tune the fuzzy sets parameters so that the total time spent in the network remains at minimum. The proposed method is tested on a stretch of a freeway network using a macroscopic traffic model in an adaptive scheme. Ozbay et al. (2003) developed an isolated feedback based ramp metering strategy that takes into account the ramp queue. In addition to the regulation of ramp input, the strategy calls for regulation of ramp queues by explicitly incorporating them into the model. This isolated rampmetering strategy is tested using PARAMICS, a microscopic traffic simulation package, on a calibrated test network located in Hayward, California. It was found that the strategy is effective in optimizing freeway traffic conditions (reduction in mean congestion duration on the freeway downstream link, mean downstream occupancy, and travel time). Coordinated Ramp Metering Strategies The bottleneck metering algorithm is a system ramp control, which includes several internal adjustments of volume reduction based on downstream bottlenecks and localized adjustments such as queue override (Jacobsen et al., 1989). At the local level, historical data is used to determine approximate volume-occupancy relationships near capacity for each ramp location. Local metering rates are then calculated to allow ramp volumes to equal the difference between the estimated capacity and the real-time upstream volume. The coordinated bottleneck algorithm is activated when the following two criteria are met: 1) A downstream bottleneck-prone section surpasses a pre-determined occupancy threshold, and 2) The zone or area of influence upstream of the bottleneck is storing vehicles. The algorithm then uses centrally assigned metering rate reductions applied to meters in the zone to reduce the number of vehicles entering the mainline by the number of vehicles stored in the bottleneck area of influence. ARMS (Advanced Real-time Metering System) consists of three operational control levels within a single algorithm: free-flow control; congestion prediction and congestion resolution (Messer, 1993). Flow is treated as a semi-static process in which traffic flow varies slowly with time, where the control decisions are based on a free flow model. Congestion prediction works to predict (and thus pre-empt) traffic flow breakdowns caused by dynamic traffic fluctuations. Traffic flow is modeled as a rapidly changing dynamic process. Integration of this control module with the free-flow control module provides for an environment in which the probability of congestion occurring is reduced. Congestion reduction is a dynamic algorithm that balances 6

9 congestion resolution time and metering rates by integrating both freeway and surface street operations. This algorithm has been successfully tested in simulation models. Wei et al. (1996) developed a coordinated metering algorithm using artificial neural networks. This algorithm is based on an Artificial Neural Network (ANN) with learning capability. It is used in an offline capacity to generate an initial, preliminary metering plan, which is used within a back-propagation algorithm to train the neural network. The roadway system is divided into control zones, and input data for the algorithm is collected at each ramp in a zone; V/C ratios upstream and downstream of the ramp and the ramp queue length on each ramp. As the metering rate for each on-ramp is affected primarily by the mainline V/C measurements near the ramp and only partially by the traffic conditions elsewhere in the zone, a partially connected neural network is used. The internal model tracks the actual traffic conditions, the implemented control strategies, and the results. This information is evaluated (expert system) and if necessary, additional self-adjustment training data is provided for the ANN system until the desired traffic condition is reached. Gettman et al. (1999) presents a multi-objective integrated large-scale optimized ramp metering system for freeway traffic management, seeking to address the interaction of the freeway system with the adjacent surface-street system by providing a method to trade-off queue growth at individual ramps in a freeway corridor. The system is composed of three primary components: area-wide metering rate coordination, predictive-cooperative real-time rate regulation, and anomaly detection/optimization scheduling. The area-wide rate coordination algorithm is based on a multi-criteria quadratic programming problem. The predictive-cooperative real-time rate regulation algorithm is a pro-active approach to local traffic-responsive control using scenario based linear programming. Re-optimization intervals of the area-wide coordination and the predictive-cooperative real-time rate regulation algorithms are scheduled by a process monitoring function based on concepts in statistical process control. The performance of the method was evaluated using a simulation test case for a typical 3-hour peak period on a realistic freeway in Phoenix, AZ in freeway average speed, total travel time, queue time, and congestion recovery time. Zhang et al. (2003) developed a new freeway ramp control objective minimizing total weighted (perceived) travel time. This new objective function is capable of balancing efficiency and equity of ramp meters, compared to a previous metering objective, which minimizes the total absolute travel time. The new objective is purely efficiency-oriented and hence produces a most efficient but least equitable solution. Consequently, a ramp control strategy BEEX (Balanced Efficiency and Equity) was developed. BEEX seeks to minimize the total weighted travel time, which involves weighting both the freeway mainline travel time and the ramp delays. A ramp metering algorithm incorporating fuzzy logic decision support was developed at the University of Washington for a number of years (Taylor et al., 1998). This algorithm was installed in early 1999 by WsDOT, controlling 15 metered ramps along I-405. The algorithm, based on fuzzy set theory, is designed to overcome some of the limitations of existing conventional ramp metering systems. In a simulation based evaluation using FRESIM and a model of the Seattle I-5 corridor, the fuzzy controller demonstrated improved robustness, prevented heavy congestion, intelligently balanced conflicting needs, and tuned easily. The 7

10 objective was to maximize total distance traveled, minimize total travel time and vehicle delay, and still maintain acceptable ramp queues. This algorithm functions on two levels, as with many of the metering algorithms available, provide both local and downstream bottleneck metering rate selection. A freeway traffic control system has been in place on the Hanshin Expressway near Kobe, Japan. The Hanshin algorithm is based on Linear Programming formulation (Yang et al., 1996). The linear algorithm maximizes the weighted sum of ramp flows. It also computes a real time capacity for each road segment. The algorithm requires a very comprehensive data collection system with detectors closely spaced on the mainline and multi-point detection on all exit/entrance ramps. To solve for metering rates, the algorithm uses both real-time and predefined system variables as well a number of tunable parameters and weighting factors for a series of ramps. The performance of the algorithm is heavily dependent on accurate origindestination data. Another coordinated ramp metering strategies, METALINE, is a coordinated generalization (using lists of multiple values, or columnar vectors, in place of single values) of ALINEA (Papageorgiou et al, 1990). The metering rate of each ramp is computed based on the change in measured occupancy of each freeway segment and the deviation of occupancy from critical occupancy for each segment that has a controlled on-ramp. This algorithm incorporates a smoothing feature from the ALINEA algorithm, preventing wide swings in metering rates between concurrent time intervals by incorporating the previous metering rate into the equation for calculating the next time interval metering rate. The sensitivity of this algorithm is also quite high, as it responds to the change in occupancy between time intervals, rather than the overall occupancy of the system, allowing more responsive operation for smaller changes in traffic flow. Chang et al. (1994) proposed a metering model for non-recurrent congestion. This algorithm uses a two-segment linear flow density model. Kalman filtering and auto-regressive moving average techniques are used for estimating link densities and ramp queue lengths from point volume and occupancy detector data and traffic system model parameters. A dynamic equation for density evolution according to the flow conservation law is formulated to describe the freeway traffic system and ramp traffic dynamics. The traffic evolution equations act as the essential constraints for optimizing metering rates. Other constraints are the lower and upper physical bounds on the mean link densities, the maximum and minimum allowable metering rates and the maximum allowable ramp queue length. Traffic flow or throughput is then solved for within the objective function using linear programming mathematics. As the successor of the ZONE metering algorithm, the Stratified Zone Ramp Metering (SZM) Strategy has been developed and deployed in the Minneapolis/Saint Paul area (Feng et al., 2005). The SZM strategy aims to maximize freeway throughput while keeping ramp waiting times below a predetermined threshold. It focused on a better determination of the minimum release rate for each ramp and its integration with the overall SZM strategy. The SZM strategy is tested in two freeway sites under various demand scenarios through a state of the art microscopic simulator. The simulation results indicate that the SZM strategy is effective in delaying and decreasing the freeway congestion as well as resulting in smoother freeway traffic flow. 8

11 In a recent study, Paramichail et al. (2010) developed a traffic- response feedback control strategy, HERO (Heuristic Ramp Metering Coordination) to coordinate local ramp metering actions in freeway networks. In the framework of HERO, ALINEA ramp metering strategy was applied to each on-ramp, the desired ramp flow was calculated, and the ramp queue was estimated. The coordination using HERO was materialized via occasional appropriate setting of minimum ramp-queue lengths that should be created and maintained at specific ramps. A pilot project of HERO has been implemented in Melbourne, Australia as a part of the Monash City in West Gate (MCW) upgrade. Wang et al. (2010) proposed an area-wide ramp metering system to improve the coordination of ramp meters for system-wide optimization and on-ramp overflow minimization. It uses the principles of a computer network congestion control strategy, which reduces certain types of congestion at a targeted freeway location through limiting on-ramp vehicle flows to a fraction of ramp demand and then additively increasing rates to avoid ramp queue spillover onto city streets. The effectiveness of this ramp metering approach has been evaluated by microscopic simulation experiments. In summary, coordinated ramp metering strategies have been suggested as more effective measures than local ramp metering when there are multiple congestion bottlenecks on the freeway, excessive ramp delays, and when the optimization of freeway and on-ramp performances requires the metering of several ramps. 9

12 2. DATA COLLECTIO Traffic data was collected from the City of Baton Rouge Regional Planning Council to reflect the forecasted origins and destinations for all on and off ramps along I-10 and I-12. In addition, geometric data was collected to build the study area network in the simulation model. A planning-level network for the Baton Rouge Metropolitan Area was provided by the Capital Regional Planning Commission (CRPC), the local MPO. The network was provided as a TransCAD file, based on the Baton Rouge Metropolitan Transportation Plan Update of December Appendix A shows the coded segments of I-10 and I-12 in TransCAD, as well each junction along the two corridors. However, only a limited number of research computers were available with this software. Therefore, the data from the provided TransCAD files needed to be extracted for use on computers without this software. TransCAD provides an export functionality which allows the user to export the proprietary file information into a variety of different formats. The database information was exported into a standard comma delimited text file (.csv) that can be easily read by Microsoft Excel or any text editor. The network node and link information was exported into a shape file (.shp). Shapes files are primarily used by ESRI ArcGIS software, but there is also other freeware available to read this file type. In addition, screenshots of all I-12 interchanges were created in Adobe pdf format which included the link ID information, in case ESRI ArcGIS was also not available. The database information provided several columns of data about each link on the network. The database, after exporting, was imported into Microsoft Excel. Most of the data were attributes assigned to each link such as: capacity, speed limit, number of lanes, facility type, and average daily traffic. Further data were the result of planning models used by the CRPC to forecast traffic patterns in the area. This data included travel time and volumes generated for the AM peak, PM peak, midday, and nighttime periods. The morning interval was defined from 6:30-9:30 AM and the evening interval from 3:30-6:00 PM. With this data, a Friction Factor Matrix was created in order to determine the origin-destination flows for the morning peak period only between on- and off-ramps as origins and destinations, respectively. The gravity model was then applied to synthesize an Origin-Destination matrix based on the estimated friction factor matrix shown in Table 1. Table 2 shows the O/D matrix after the first iteration with very large errors in the attractions. A total of 14 iterations were required to reduce all errors in the attractions below 1%. The final O/D matrix, shown in Table 3, was then used in the simulation model to predict the network performance with and without ramp metering strategies in the target year. 10

13 Table 1: Friction Factor Matrix D Labels Off1 Off2 Off3 Off4 Off5 Off6 Off7E Off8E Off9E A23 Off20 Off22 Off24 A60 Off12 Off11 Off10 Off9W Off8W Off7a Off7W A24 Off17 Off15a Off15 O Labels Flows A On On On On On On On13E On13a On13b On14E A On On On On On On14W On13W On On A On On On

14 Table 2: First Iteration Origin/Destination Matrix for I-10/I-12 Corridors for the Period of 6:30-9:30 AM Error D Labels Off1 Off2 Off3 Off4 Off5 Off6 Off7E Off8E Off9E A23 Off20 Off22 Off24 A60 Off12 Off11 Off10 Off9W Off8W Off7a Off7W A24 Off17 Off15a Off15 O Labels Flows A On On On On On On On13E On13a On13b On14E A On On On On On On14W On13W On On A On On On

15 Table 3: Final Origin/Destination Matrix for I-10/I-12 Corridors for the Period of 6:30-9:30 AM Error (%) D Labels Off1 Off2 Off3 Off4 Off5 Off6 Off7E Off8E Off9E A23 Off20 Off22 Off24 A60 Off12 Off11 Off10 Off9W Off8W Off7a Off7W A24 Off17 Off15a Off15 O Labels Flows A On On On On On On On13E On13a On13b On14E A On On On On On On14W On13W On On A On On On

16 3. METHODOLOGY ETWORK DESCRIPTIO The simulation platform used in this study is VISSIM version 5, which is a microscopic, behavior based and time step simulation model. VISSIM is a valuable tool for the analysis of various devices used in transportation engineering because of its ability to model traffic behavior at a microscopic level under varying constraints such as lane configuration, traffic composition and traffic signals. The freeway corridors of I-10 and I-12 were coded in VISSIM using links and nodes. Each link represents a single approach and each connector serves to model junctions. Each link must have a connector that bonds the next consecutive link to the previous one. These tasks were repeated until both corridors were completely constructed for I-10 and I-12 within the Baton Rouge area. Figure 1 shows a snapshot of both corridors as coded in VISSIM. Figure 1: VISSIM Coded etwork for I-10 and I-12 After the construction of both corridors, routes were created from every specific entrance point (on-ramp) to all possible exit points (off-ramps) for both eastbound and westbound directions within the simulation model. Each route began at a routing decision point and ended at one or more destination points. For each designated route, a number of trips were assigned based on the final O/D matrix explained earlier. 14

17 SIMULATIO EXPERIME TS In order to examine the network performance with and without the implementation of ramp metering strategies, two simulation scenarios were created, one with ramp meters and one without ramp meters. For the ramp meter scenario, a ramp meter controller was added for each on-ramp along both corridors in both directions. Also, signal heads were installed at every on ramp to represent each ramp meter. A set of detectors was also attached to each signal head. One detector was placed at the location of the signal head and another one shortly behind signal head. Other detectors were added on each lane of the mainline to adjust the ramp meter flow rate based on the current lane occupancy detected on the mainline. Each set of detectors was identified with its reference signal head by a two digit number system where the tens digit was the signal controller and the ones digit was the detector numbers. For the control corridor, no signal heads or detectors were created, as no ramp meters would be used. Vehicle Actuated Programming (VAP) was used as the signal state generator. With this setting, user controlled signal logic was actuated. SIMULATIO RU S For each of the two simulation scenarios, 25 simulation runs were used to evaluate the network performance. Each run was generated from a random seed. The network was simulated for one hour, in addition to a 15-minute warm up period. A set of network-level performance measures was also identified as follows: Average delay time per vehicle [s], All Vehicle Types Average number of stops per vehicles, All Vehicle Types Average speed [mph], All Vehicle Types Average stopped delay per vehicle [s], All Vehicle Types Total delay time [h], All Vehicle Types Total Distance Traveled [mi], All Vehicle Types Number of Stops, All Vehicle Types Total stopped delay [h], All Vehicle Types Total travel time [h], All Vehicle Types 15

18 4. SIMULATIO A D A ALYSIS OF RESULTS This section presents the results of the simulation runs for each of the two scenarios. In order to account for randomness in driving behavior, a total of 25 simulation runs were generated with VISSIM. The results of each simulation run are summarized in Table 4 and Table 5 for the scenario with activated ramp meters and in Table 6 and Table 7 for the scenario without ramp meters. For each simulation run, the results are shown for each of the selected performance measures. Graphical illustration and statistical analyses to compare the network performance for both scenarios are presented next. Avg. # Stops per vehicles, All Vehicle Types Table 4: Results of Simulation Runs for Metered Traffic (Part I) Avg. Stopped delay per vehicle [s], All Vehicle Types Total delay time [h], All Vehicle Types Total Distance Traveled [mi], All Vehicle Types # Stops, All Vehicle Types Run

19 Table 5: Results of Simulation Runs for Metered Traffic (Part II) Total stopped delay [h], All Vehicle Types Avg Delay time per vehicle [s], All Vehicle Types Total travel time [h], All Vehicle Types Avg Speed [mph], Run All Vehicle Types

20 Table 6: Results of Simulation Runs for on-metered Traffic (Part I) Average delay time per vehicle [s], All Vehicle Types Average number of stops per vehicles, All Vehicle Types Average speed [mph], All Vehicle Types Average stopped delay per vehicle [s], All Vehicle Types Total delay time [h], All Vehicle Types Run

21 Table 7: Results of Simulation Runs for on-metered Traffic (Part II) Total Distance Traveled [mi], All Vehicle Types 19 Total stopped delay [h], All Vehicle Types Total travel time [h], All Vehicle Types Number of Stops, Run All Vehicle Types COMPARATIVE EVALUATIO This section compares the identified performance measures for both metered and non-metered traffic using graphical illustrations of each measure for the 25 simulation runs. Average umber of Stops per Vehicle The average number of stops per vehicle describes driving in congested conditions. High number of stops implies frequent stops as often observed in stop and go driving conditions. Figure 2 shows that the average number of stops for non-metered traffic falls in the range of 30 to 31 stops per vehicle. For metered traffic, however, the average falls in the range of stops. The difference implies relatively smoother traffic and fewer stops when ramp metering strategies are implemented.

22 Figure 2: Average umber of Stops per Vehicle Total Delay Time The total delay time describes the total delay experienced by all vehicles in the network. Figure 3 shows that the total delay experienced by all vehicles with ramp metering implementation is nearly 750 hours less than that experienced without ramp meters. This is equivalent to a saving of nearly 18% in delay when ramp metering strategies are implemented. Figure 3: Total Delay Time in Vehicle Hours 20

23 Total Stopped Delay The total stopped delay describes s the total delay experienced by vehicles in stopped conditions. This excludes acceleration and deceleration delays. Figure 4 shows that there is nearly a reduction of 200 hours of stopped delay when ramp metering strategies are implemented. This is equivalent to a saving of 17% as a result of using ramp meters. Average Delay Time per Vehicle Figure 4: Total Stopped Delay (Vehicle Hours) The average delay time per vehicle is the total delay divided by the total number of vehicles in the network. Figure 5 shows that t for all simulation runs, the average delay for non-metered traffic exceeded that for metered traffic. The difference amounts to nearly 9% saving in average delay per vehicle and implies that ramp metering strategies would produce savings in average vehicular delays. 21

24 Figure 5: Average Delay Time per Vehicle (Seconds) Average Stopped Delay per Vehicle Similar to the average delay per vehicle, the average stopped delay is the total stopped delay divided by the total number of vehicles. Figure 6 shows a relatively higher variation in the average stopped delay for metered traffic compared to non-metered traffic. However, for almost all simulation runs the average stopped delay for non-metered traffic exceeds that for metered traffic. Overall, a reduction of nearly 6% is observed for all simulation runs. This also illustrates the benefit of ramp metering strategies. Figure 6: Average Stopped Delay per Vehicle (Seconds) 22

25 Total Travel Time Another performance measure is the total travel time spent by all vehicles in the network. Figure 7 clearly shows that the travel time for metered traffic is lower than that for non-meterea reduction of nearly 13% in total travel time was achieved by ramp traffic. On average, metering. Average Speed Figure 7: Total Travel Time (Hours) In terms of average speed, although a difference in the order of nearly one mph was observed, the trend is consistent with the previous results and shows that average speed for metered traffic is higher than that for non-metered traffic. Figure 8: Average Speed (mph) 23

26 STATISTICAL A ALYSIS This section presents the statistical analysis used to compare the traffic performance for metered and non-metered traffic on the study section. Basic descriptive statistics is presented first, followed by tests of hypothesis. All tests were performed in SAS. Descriptive Statistics The basic descriptive statistics for metered and non-metered scenarios are shown in Table 8 and Table 9, respectively. The statistics include the sample size, the mean, the standard deviation, the minimum and maximum values. Table 8: Basic Descriptive Statistics for Metered Corridors Variable Mean Std Dev Minimum Maximum Average number of stops per vehicles Average stopped delay per vehicle [s] Total delay time [h] Total distance traveled [mi] Total number of stops Total number of vehicles in the network Number of vehicles that have left the network Total stopped delay [h] Average delay time per vehicle [s] Average speed [mph] Total travel time [h] Table 9: Basic Descriptive Statistics for on-metered Corridors Variable N Mean Std Dev Minimum Maximum Average number of stops per vehicles Average stopped delay per vehicle [s] Total delay time [h] Total distance traveled [mi] Total number of stops Total number of vehicles in the network Number of vehicles that have left the network Total stopped delay [h] Average delay time per vehicle [s] Average speed [mph] Total travel time [h] Tests of Hypotheses The statistical tool used for the analyses of results was the Student s t-test for two independent population means with unknown variances. The population variances were first tested to confirm whether they were equal or not; so as to determine whether to perform a pooled t-test or Satterthwaite t-test, respectively. In addition, Fisher s Least Significant Difference (LSD) method was used to obtain a 95% confidence interval estimate of the difference between the two 24

27 population means, and results compared to that obtained using the t-test. For each variable, the following hypotheses were tested: : The mean population values between the Non-metered and Metered Corridors are the same : The mean population values between the Non-metered and Metered Corridors are different Average umber of Stops per Vehicle The values of skewness and kurtosis obtained were and , respectively. Since these are close to zero, it is assumed that this variable is normally distributed. If and denote the average number of stops per vehicle for the metered and non-metered corridors, respectively, the following hypotheses were tested: : = 0 : 0 (no difference exists) (difference exists) A test of variances concluded both populations have equal variances, and therefore, a pooled t- test analysis was performed on the population means. This resulted in a p-value of <0.0001, which is less than the 0.05 level of significance used. It can therefore be concluded at the 0.05 level of significance that a difference exists between the average number of stops per vehicle within the metered and non-metered corridors. In particular, since =27.61 and =30.69, it can be concluded that the average number of stops per vehicle is greater within the nonmetered corridor than it is within the metered corridor. In addition, Fisher s LSD produced a 95% confidence interval estimate of [-3.35, -2.88] for. Since every point in this interval is less than zero, it can be concluded at the 95% significance level that the average number of stops within the non-metered corridor is greater than the average number of stops within the metered corridor. This agrees with the t-test analysis. Total Delay Time The values of skewness and kurtosis obtained were and , respectively. Since these are close to zero, it is assumed that this variable is normally distributed. If and denote the total delay time in hours for the metered and non-metered corridors respectively, the following hypotheses were tested: : = 0 : 0 (no difference exists) (difference exists) A test of variances concluded both populations have equal variances therefore a pooled t-test analysis was performed on the population means. This resulted in a p-value of < which is less than the 0.05 level of significance used. It can therefore be concluded at the 0.05 level of significance that a difference exists between the total delay time within the metered and non- 25

28 metered corridors. In particular, since =3, and =4,210.40, it can be concluded that the total delay time in hours is greater within the non-metered corridor than within the metered corridor. In addition, Fisher s LSD produced a 95% confidence interval estimate of [ , ] for. Since every point in this interval is significantly less than zero, it can be concluded at the 95% significance level that the total delay time in hours within the non-metered corridor is significantly greater than the total delay time in hours within the metered corridor. This agrees with the t-test analysis. Total Stopped Delay The values of skewness and kurtosis obtained were and , respectively. Since these are close to zero, it is assumed that this variable is normally distributed. If and denote the total stopped delay in hours for the metered and non-metered corridors respectively, the following hypotheses were tested: : = 0 : 0 (no difference exists) (difference exists) A test of variances concluded both populations have equal variances therefore a pooled t-test analysis was performed on the population means. This resulted in a p-value of < which is less than the 0.05 level of significance used. It can therefore be concluded at the 0.05 level of significance that a difference exists between the total stopped delay time within the metered and non-metered corridors. In particular, since =1, and =1,187.10, it can be concluded that the total stopped delay in hours is greater within the non-metered corridor than within the metered corridor. In addition, Fisher s LSD produced a 95% confidence interval estimate of [ , ] for. Since every point in this interval is significantly less than zero, it can be concluded at the 95% significance level that the total stopped delay within the non-metered corridor is significantly greater than the total stopped delay within the metered corridor. This agrees with the t-test analysis. Average Delay Time per Vehicle The values of skewness and kurtosis obtained were and , respectively. Since these are close to zero, it is assumed that this variable is normally distributed. If and denoted the average delay time in seconds per vehicle for the metered and non-metered corridors respectively, the following hypotheses were tested: : = 0 : 0 (no difference exists) (difference exists) 26

29 A test of variances concluded both populations have equal variances therefore a pooled t-test analysis was performed on the population means. This resulted in a p-value of < which is less than the 0.05 level of significance used. It can therefore be concluded at the 0.05 level of significance that a difference exists between the average delay time within the metered and nonmetered corridors. In particular, since = and =496.93, it can be concluded that the average delay time in seconds is greater within the non-metered corridor than within the metered corridor. In addition, Fisher s LSD produced a 95% confidence interval estimate of [-48.35, ] for. Since every point in this interval is less than zero, it can be concluded at the 95% significance level that the average delay time in seconds per vehicle within the non-metered corridor is greater than the average delay time in seconds per vehicle within the metered corridor. Average Stopped Delay per Vehicle The values of skewness and kurtosis obtained were and , respectively. Since these are close to zero, it is assumed that this variable is normally distributed. If and denote the average stopped delay time in seconds per vehicle for the metered and non-metered corridors respectively, the following hypotheses were tested: : = 0 : 0 (no difference exists) (difference exists) A test of variances concluded the populations have unequal variances therefore a Satterthwaite t- test analysis was performed on the population means. This resulted in a p-value of < which is less than the 0.05 level of significance used. It can therefore be concluded at the 0.05 level of significance that a difference exists between the stopped delay time per vehicle within the metered and non-metered corridors. In particular, since = and =139.93, it can be concluded that the stopped delay time in seconds per vehicle is greater within the nonmetered corridor than within the metered corridor. In addition, Fisher s LSD produced a 95% confidence interval estimate of [-10.11, -5.02] for. Since every point in this interval is less than zero, it can be concluded at the 95% significance level that the average stopped delay time in seconds per vehicle within the nonmetered corridor is greater than the average stopped delay time in seconds per vehicle within the metered corridor. This agrees with the t-test analysis. Total Travel Time The values of skewness and kurtosis obtained were and , respectively. Since these are close to zero, it is assumed that this variable is normally distributed. If and denoted the total travel time in hours for the metered and non-metered corridors respectively, the following hypotheses were tested: : = 0 (no difference exists) 27

30 : 0 (difference exists) A test of variances concluded both populations have equal variances therefore a pooled t-test analysis was performed on the population means. This resulted in a p-value of < which is less than the 0.05 level of significance used. It can therefore be concluded at the 0.05 level of significance that a difference exists between the total travel time within the metered and nonmetered corridors. In particular, since =5, and =6,066.90, it can be concluded that the total delay travel time in hours is greater within the non-metered corridor than within the metered corridor. In addition, Fisher s LSD produced a 95% confidence interval estimate of [ , ] for. Since every point in this interval is significantly less than zero, it can be concluded at the 95% significance level that the total travel time in hours within the non-metered corridor is significantly greater than the total travel time in hours within the metered corridor. This agrees with the t-test analysis. Average Speed The values of skewness and kurtosis obtained were and , respectively. Since these are close to zero, it is assumed that this variable is normally distributed. If and denoted the average speed per vehicle in mph for the metered and non-metered corridors respectively, the following hypotheses were tested: : = 0 : 0 (no difference exists) (difference exists) A test of variances concluded both populations have equal variances therefore a pooled t-test analysis was performed on the population means. This resulted in a p-value of < which is less than the 0.05 level of significance used. It can therefore be concluded at the 0.05 level of significance that a difference exists between the average speeds within the metered and nonmetered corridors. In particular, since =10.88 and =9.99, it can be concluded that the average speed in mph is greater within the metered corridor than within the non-metered corridor. In addition, Fisher s LSD produced a 95% confidence interval estimate of [0.83, 0.96] for. Since every point in this interval is greater than zero, it can be concluded at the 95% significance level that the average speed per vehicle in mph per vehicle within the metered corridor is greater than the average speed per vehicle in mph within the non-metered corridor. This agrees with the t-test analysis. 28

31 5. CO CLUSIO S SUMMARY The primary goal of this research was to lay a foundation for the application and implementation of integrated corridor management strategies to reduce congestion on the freeway and arterial systems in Baton Rouge. Due to the exponential increase in population and travel demand, Baton Rouge may potentially benefit from integrated corridor management strategies such as ramp metering which essentially targets congestion mitigation. This study applied ramp metering strategies on the two corridors of I-10 and I-12 within the city of Baton Rouge in order to determine their effectiveness. This was achieved by simulating both corridors with and without ramp metering at the microscopic level using the forecasted traffic demand in the year Traffic data was collected from the City of Baton Rouge Regional Planning Council to reflect the forecasted origins and destinations for all on and off ramps along I-10 and I-12. In addition, geometric data was collected to build the study area network in the simulation model. With this data, a Friction Factor Matrix was created in order to determine the origin-destination flows for the morning peak period only between on- and off-ramps as origins and destinations, respectively. The gravity model was then applied to synthesize an Origin-Destination matrix based on the estimated friction factor matrix. A total of 14 iterations were required to reduce all errors in the attractions below 1%. The simulation platform used in this study was VISSIM version 5, which is a microscopic, behavior based and time step simulation model. The freeway corridors of I-10 and I-12 were coded in VISSIM using links and nodes. In order to examine the network performance with and without the implementation of ramp metering strategies, two simulation scenarios were created, one with ramp meters and one without ramp meters. For each of the two simulation scenarios, 25 simulation runs were used to evaluate the network performance. Each run was generated from a random seed. The network was simulated for one hour, in addition to a 15-minute warm up period. The following is the set of network-level performance measures used in the analysis: Average delay time per vehicle [s], All Vehicle Types Average number of stops per vehicles, All Vehicle Types Average speed [mph], All Vehicle Types Average stopped delay per vehicle [s], All Vehicle Types Total delay time [h], All Vehicle Types Total Distance Traveled [mi], All Vehicle Types Number of Stops, All Vehicle Types Total stopped delay [h], All Vehicle Types Total travel time [h], All Vehicle Types 29

32 FI DI GS The comparative evaluation of both scenarios (with and without ramp metering) shows a statistically significant improvement in the corridor performance when ramp metering strategies are implemented. The statistical analysis using the Student s t-test for two independent samples with unknown variances showed consistently that the means were significantly different at 95% confidence level. A test of variances was also conducted and concluded that both populations had equal variances, and therefore, a pooled t-test analysis was conducted. Also, Fisher s Least Significant Difference (LSD) method was applied on the difference between the two population means at 95% confidence level. The statistical analysis shows that: The average number of stops within the non-metered corridor is greater than the average number of stops within the metered corridor. The total delay time in hours is greater within the non-metered corridor than within the metered corridor. The total stopped delay in hours is greater within the non-metered corridor than within the metered corridor. The average delay time in seconds is greater within the non-metered corridor than within the metered corridor. The stopped delay time in seconds per vehicle is greater within the non-metered corridor than within the metered corridor. The total delay travel time in hours is greater within the non-metered corridor than within the metered corridor. The average speed in mph is greater within the metered corridor than within the nonmetered corridor. Based on the simulation results, the study recommends the use of ramp metering on both segments of I-10 and I-12. It should be noted, however, that additional research is still required to optimize the ramp metering parameters to further improve the corridor performance. Another potential improvement could be achieved through coordination of ramp meters along the corridor. 30

33 Chang, 6. LIST OF REFERE CES 1. H G. J. Wu, and S. Cohen Integrated Real-Time Ramp Metering Model for Nonrecurrent Congestion: Framework and Preliminary Results. Transportation Research Record 1446, Feng, B., J. Hourdos, and P. Michalopoulos, Improving Minnesota s Stratified Ramp Control Strategy, Transportation Research Board 84 th Annual Meeting, Washington D.C., Gettman D. M., S. Corporation, and G. Systems, A multi-objective integrated large-scale optimized ramp metering control system (MILOS) for freeway traffic management. Transportation Research Board 78 th Annual Meeting, Washington D.C., Ghods, A. H., A. R.Kian, and M.Tabibi, Adaptive Freeway Ramp Metering and Variable Speed Limit Control: A Genetic-Fuzzy Approach. IEEE Intelligent Transportation Systems Magazine, Volume 1, Issue 1, pp , Jacobsen, L, K. Henry, and O. Mahyar, Real-Time Metering Algorithm for Centralized Control. Transportation Research Record 1232, TRB, Jia, Z., C. Chen, B. Coifman, and P. Varaiya. The PeMS algorithms for accurate, real time estimates of g-factors and speeds from single-loop detectors. In Proc. 4th IEEE Conference on Intelligent Transportation Systems, Oakland, California, 2001, pp Masher, D.P., D.W. Ross, P.J. Wong, P.L. Tuan, H.M. Zeidler, and S. Petracek. Guidelines for Design and Operation of Ramp Control Systems. Stanford Research Institute, Menlo Park, California, Messer, C. Advanced Freeway System Ramp Metering Strategies for Texas; Texas Department of Transportation Report Number , Ozbay, K, I. Yasar, and P. Kachroo, Modeling and PARAMICS Based Evaluation of New Local Freeway Ramp-Metering Strategy that Takes Ramp Queues into Account, Transportation Research Board 83 th Annual Meeting, Washington D.C., Papageorgiou, M., H. Hadj-Salem, and F. Middelham, ALINEA Local Ramp Metering Summary of Field Results. Transportation Research Record 1603, TRB, Papageorgiou, M., Blosseville, J. M., and Hadj Salem, H. Modeling and Real-time Control of Traffic Flow on the Southern Part of Boulevard Peripherique in Paris: Part II: Coordinated On-ramp Metering, Transportation Research Vol. 24A, No. 5, pp , Papamichail, I., M. Papageorgiou, V. Vong, and J. Gaffney, HERO Coordinated Ramp Metering Implementation at the Monash Freeway. Transportation Research Board 89 th Annual Meeting, Washington D.C., Smaragdis, E. and M. Papageorgiou, A Series of New Local Ramp Metering Strategies. Transportation Research Board 82 th Annual Meeting, Washington D.C., Taylor, C., D. Meldrum, and L. Jacobson, Fuzzy Ramp Metering - Design Overview and Simulation Results, Transportation Research Record 1634, Washington Thompson, N. and S. Greene, Ramp Metering for the 21st Century: Minnesota's Experience. Proceedings of America 7th Annual Meeting, April

34 16. Wang, Y., K. A., Perrine, and Y., Lao, Developing an Area-Wide System for Coordinated Ramp Meter Control. Transportation Research Board 89 th Annual Meeting, Washington D.C., Wei, C. and K. Wu Applying an Artificial Neural Network Model To Freeway Ramp Metering Control, Transportation Planning Journal, Vol. 25 No. 3; Yang, X., Y. Iida, N Uno, and P. Yang; Dynamic Traffic Control System for Urban Expressway with Constraint of Off-Ramp Queue Length ; ITS World Conference Proceedings, Seoul Zhang, L. and D. Levinson, Balancing Efficiency and Equity of Ramp Meters. Transportation Research Board 82 th Annual Meeting, Washington D.C.,

35 APPE DIX A I-10/I-12 STUDY SEGME TS A D FREEWAY JU CTIO S 33

36 Figure 9: Study Segments of I-10 and I-12 34

37 Figure 10: Airline Junction 35

38 Figure 11: Bluebonnet Junction 36

39 Figure 12: College Dr. Junction 37

40 Figure 13: Essen Junction on I-10 38

41 Figure 14: Essen Junction on I-12 39

42 Figure 15: Frost Livingston Junction 40

43 Figure 16: Gonzales Junction 41

44 Figure 17: Highland Junction 42

45 Figure 18: Jefferson at Drusilla Junction 43

46 Figure 19: Juban Junction 44

47 Figure 20: LA 22 Junction on I-10 45

48 Figure 21: LA 44 Junction on I-10 46

49 Figure 22: Millerville Junction 47

50 Figure 23: icholson Highland Junction 48

51 Figure 24: O eil Junction 49

52 Figure 25: Perkins/Acadian Junction 50

53 Figure 26: Picue Lane Junction 51

54 Figure 27: Port Allen Junction 52

55 Figure 28: Prairieville Junction 53