Jack Klodzinski

EVALUATING THE IMPROVEMENTS IN TRAFFIC OPERATIONS AT A REAL-LIFE TOLL PLAZA WITH ELECTRONIC TOLL COLLECTION

This paper summarizes the improvements in traffic operations at the Electronic Toll Collection plazas of the Orlando-Orange County Expressway Authority (OOCEA). Automatic Vehicle Identification (AVI) was installed on all lanes of ten OOCEA plazas, with one fully dedicated AVI lane at each plaza. The results indicate that the measured throughput of the dedicated AVI lane has increased by 154%. Significant improvements were observed at the dedicated AVI lane including a 194% potential increase in capacity, average savings of 5 seconds in service time per vehicle, a reduction of 1 minute in average queueing delay and 2.5 minutes in maximum queueing delay per vehicle, and average savings of 8.5 vehicle-hours in total queueing delay per peak hour. Also, variability in the inter-vehicle time has been significantly reduced in the dedicated AVI lane. Capacity, inter-vehicle time, and service times of the mixed lanes did not change significantly because of AVI. However, arrivals have shifted to the dedicated AVI l...

Evaluation of Two Modeling Methods for Generating Heavy-Truck Trips at an Intermodal Facility by Using Vessel Freight Data

A methodology for building a truck trip generation model by use of artificial neural networks from vessel freight data has been developed and successfully applied to five Florida seaports. The backpropagation neural network (BPNN) algorithm was used in the design. Although the methodology was sound, a new model had to be developed for each of these intermodal facilities. Lead and lag variables were necessary input variables for most models to account for commodities stored on port property before export or pickup after import. Other modeling techniques were researched, and a fully recurrent neural network (FRNN) trained by the real-time recurrent learning algorithm was selected to develop a model for Port Canaveral and compare with a BPNN model. FRNN is dynamic in nature and was found to relate to the storage time of the commodities to truck trip generation. A developed Port Canaveral BPNN model was successfully validated at the 95% confidence level with collected field data. It was applied to conduct a s...

Proposed Level-of-Service Methodology for Toll Plazas

A proposed macroscopic methodology for measuring the level of service (LOS) of toll plazas has been developed using delay as the measure of effectiveness (MOE). On the basis of field research and data analyses, the 85th percentile of the cumulative individual vehicular delay was found to be the most comprehensive measure for evaluating the LOS at a toll plaza. Other MOEs were examined but found to be less flexible with different plaza configurations and lane payment types. More than 55,000 individual vehicular records from three mainline toll plazas in Orlando, Florida, representing eight different plaza configurations with varied percentages of electronic toll collection (ETC) usage were used to validate the methodology. TPSIM, a toll plaza simulation model, was used to produce an additional 49 scenarios representing the three plazas with varied percentages of ETC usage and 21 additional plaza configurations. Service time was examined to determine the level at which a driver begins to feel discomfort and inconvenience at a toll plaza. An LOS hierarchy was established based on the conclusions of this analysis, feedback from professionals, and reference to the Highway Capacity Manual (HCM) 2000. The 85th delay percentile graphs from each of the plaza analysis results for LOS values were also observed to be similar.

Methodology for Modeling a Road Network with High Truck Volumes Generated by Vessel Freight Activity from an Intermodal Facility

An innovative methodology has been developed for analyzing freight movement on local road networks by merging previously developed truck trip generation models using artificial neural networks (ANNs) and a microscopic network simulation model. Through computer simulation, this methodology comprehensively analyzes a seaport considered a special generator of heavy truck traffic and an adjacent road network that includes identified intermodal routes that connect to a seaport. Truck traffic from the seaport is initially modeled with ANNs using vessel freight activity at the seaport. These ANN models have been incorporated into the methodology to provide accurate truck and total traffic volumes for modeling the networks. This methodology was successfully tested with two network microscopic simulation models. Transferability was successfully tested with two seaports with different characteristics. Three months of field data from each port and selected locations on the networks were used in calibration and validation. Both models were successfully validated and showed no statistically significant difference between the field and model output data. This methodology can be used to evaluate local port networks to manage traffic efficiently during heavy congestion or investigate forecasted port growth. These networks can also be used for information technology applications such as incident management or alternative route choice. The ability to develop a network that includes significant truck volumes generated by freight activity is a useful tool especially for engineers and planners involved in intermodal transportation analysis.

Transferability of an intermodal freight transportation forecasting model to major Florida seaports

Seaports are important economic generators, and identifying necessary infrastructure improvements is essential to accommodate potential growth at these intermodal facilities. The ability of heavy trucks to access a port’s freight terminals is one such operational improvement that needs to be addressed. Freight activity from Florida’s major seaports generates more than 10,000 trucks per day. Efficient accessibility to freight terminals and storage facilities at ports can be provided by identifying needed improvements in transportation operations and by developing truck trip generation models to forecast truck trips in and out of the ports. In July 2001 an artificial neural network (ANN) model was developed successfully by using backpropagation techniques to simulate the transportation of freight by heavy trucks generated from an intermodal activity center such as a seaport. The general methodology for developing the model was applied to the Port of Tampa and Port Canaveral in Florida to test the transferability of the ANN modeling technique. From daily vessel freight data, models for both ports were developed successfully and validated at the 95% confidence level with data collected from the field. The validated models were executed for short-term forecasts of truck trips at both ports. Port of Tampa was forecast to have a 0.31% average annual decrease in heavy trucks, attributed to the decreasing trend in bulk commodity shipments. A 5.07% average annual increase was forecast for Port Canaveral, which correlated with historical trends and future estimates for freight activity. Output from these models can be directly used as truck variable inputs to route assignment models used by local and regional government agencies.

Evaluation of Toll Plaza Performance After Addition of Express Toll Lanes at Mainline Toll Plaza

Conversion of a conventional toll plaza to an open-road tolling design with express electronic toll collection (ETC) lanes improved operating conditions for cash and ETC customers during the peak a.m. and p.m. periods. Data were analyzed before and after the plazas renovation to evaluate its performance. Vehicle delay was not excessive at the plaza during the before-study analysis except for heavy demand during the p.m. peak southbound partly because of a reduction in available service lanes. The typical average delay per cash-paying vehicle was 14 s. After the redesign, the analysis showed a reduction in the average delay per vehicle of 7 s for the automatic coin machine (ACM) lanes and 8 s for the manual cash payment lanes. The average delay per lane per hour was reduced by over 4,560 s for the ACM lanes and 3,360 s for the manual payment lanes. Reduction in total plaza delay for cash customers averaged more than 20 s/veh in the after study. ETC use in the cash payment lanes previously averaged over 10% and dropped below 3% during the a.m. peak. For the exception of the ACM lanes during the p.m. peak northbound, the overall ETC use in the cash lanes was now less than 5%, showing the benefit recognized by ETC users at the plaza from express lanes. During the before study, the plaza capacity in the peak direction during a peak hour was estimated at a maximum of 5,624 vehicles per hour. After implementation of open-road tolling, the estimated capacity increased by 43.8%.

TRANSFERABILITY OF A STOCHASTIC TOLL PLAZA COMPUTER MODEL

Transferability of the TPSIM toll plaza microsimulation computer model to other toll plazas was explored with 3 days of data collected between 1994 and 2000 at Dean Plaza, an Orlando–Orange County Expressway Authority (OOCEA) toll plaza. TPSIM is a discrete-event stochastic microscopic simulation model that has been previously validated at the 95% confidence level with data from OOCEA’s busiest toll plaza, Holland East Plaza. This model has the capability to accurately model virtually any toll plaza scenario. To apply this model to another toll plaza, a calibration procedure is necessary. To identify the modifications necessary to the input parameters, over 400 simulation runs with the 3 selected days from Dean Plaza were compared and statistically analyzed for accuracy. An experimental design was outlined and followed to accurately account for each trial simulation run during the calibration process. The service time was determined to have the most significant impact on the simulation model in much the same manner as it does in the field. The measures of effectiveness (MOEs) chosen to evaluate TPSIM were throughput, average queuing delay, maximum queuing delay, and total queuing delay. At the 95% confidence level, no significant difference was found in the comparison of field and simulated MOE values for all 3 days. Comparison of the Holland East and Dean Plaza results indicates that the TPSIM simulation model is transferable to another toll plaza with different configurations.

Evaluating Effects of Toll Strategies on Route Diversion and Travel Times for Origin–Destination Pairs in a Regional Transportation Network

Congestion on freeway facilities is a growing menace. Interstate 4 (I-4) in the Central Florida region has been experiencing delays during peak hour; this has warranted research on traffic management strategies. The public, through the media, had proposed removing tolls on state toll roads to divert traffic from I-4. A microsimulation model, Paramics, was used to examine the potential impact of this proposal. SR-417 is a relatively uncongested toll road alternative to I-4. SR-528 is the east–west toll road connecting SR-417 and I-4. Commuters on SR-417 have to travel 15 mi longer and pay

DEVELOPMENT OF A JAVA APPLET FOR GENERATING TRUCK TRIPS FROM FREIGHT DATA

5 compared with no monetary cost on I-4 for the same trip. The public and politicians are reluctant to toll I-4 to relieve congestion. The results from the simulation indicated that under recurring congestion conditions on I-4, removing tolls on SR-417 and SR-528 would not divert enough traffic from I-4 because of the 15-mi advantage. Under incident and lane closure scenarios on I-4 with toll reduction on SR-417 and SR-528, the travel time would increase on I-4. This result would prompt some diversion, with volumes and travel times increasing on SR-417. It was concluded that the amount of traffic that would be diverted from I-4 to the toll roads would not significantly relieve congestion on I-4. When specific origin–destination pairs were analyzed, average travel time savings on I-4 were only around 5 min. It was concluded that contrary to the media and public perception, toll reduction would only have a minimum impact on reducing I-4 congestion.

Methods for Quantitative Risk Analysis for Travel Demand Model Forecasts

As freight transportation becomes a more significant concern, the ability to estimate accurately truck trips generated by freight activity at an intermodal facility is important for transportation engineering and planning. Truck trip generation models that used vessel freight data were developed by the University of Central Florida Transportation Systems Institute and have been statistically proved to determine accurately the number of trucks generated at a seaport. To apply these previously developed models more efficiently, a Java applet was developed to execute a selected artificial neural network (ANN) port model and a trainable ANN port model. This applet provides the user with an easy-to-understand interface for entering data and executing models. The two models developed were the Port Everglades ANN model and a hybrid of the Port Everglades model that allowed the user to retrain the ANN model before execution for desired data. Although the trainable model requires additional retraining with new data, the added complexity comes with the benefit of producing a network with a higher degree of accuracy. This applet equips the user with two models and therefore has expanded the capabilities of the previously limited Port Everglades ANN truck trip generation model. The developed Java applet can be expanded to include more ANN models and thus more flexibility, depending on the type of freight data available. This successful adaptation of the ANN model into a Java applet sets the foundation for further applications.