AN INTERSECTION TRAFFIC DATA COLLECTION DEVICE UTILIZING LOGGING CAPABILITIES OF TRAFFIC CONTROLLERS AND CURRENT TRAFFIC SENSORS.

AN INTERSECTION TRAFFIC DATA COLLECTION DEVICE UTILIZING LOGGING CAPABILITIES OF TRAFFIC CONTROLLERS AND CURRENT TRAFFIC SENSORS Final Report November 2008 UI Budget KLK134 NIATT Report Number N08-13 Prepared by National Institute for Advanced Transportation Technology University of Idaho Ahmed Abdel-Rahim Brian k. Johnson

DISCLAIMER The contents of this report reflect the views of the authors, who are responsible for the facts and the accuracy of the information presented herein. This document is disseminated under the sponsorship of the 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.

1. Report No. 2. Government Accession No. 4. Title and Subtitle An Intersection Traffic Data Collection Device Utilizing Logging Capabilities of Traffic Controllers and Current Traffic Sensors 7. Author(s) Abdel-Rahim, Dr. Ahmed; Johnson, Dr. Brian 3. Recipient s Catalog No. 5. Report Date November 2008 6. Performing Organization Code KLK134 8. Performing Organization Report No. N08-13 9. Performing Organization Name and Address 10. Work Unit No. (TRAIS) National Institute for Advanced Transportation Technology University of Idaho PO Box 440901; 115 Engineering Physics Building Moscow, ID 83844-0901 12. Sponsoring Agency Name and Address US Department of Transportation Research and Special Programs Administration 400 7th Street SW Washington, DC 20509-0001 11. Contract or Grant No. DTRS98-G-0027 13. Type of Report and Period Covered Final Report: November 2006 August 2008 14. Sponsoring Agency Code USDOT/RSPA/DIR-1 15. Supplementary Notes: 16. Abstract The project presents a high-resolution data logging device that can be used in real-time traffic monitoring at signalized intersections. The data logging device can be connected to traffic cabinets using different connection modes. The data logging device logs the status of all input and output communication channels and updates their status continuously. This data can be accessed remotely through an Ethernet port over IP based communication. The data logging device presented in this project provides an opportunity for high-resolution real-time performance monitoring of intersection operations. The project presents two applications in which the data logging device was used to monitor intersection performance. In the first application, the device was used to plot continuous time-occupancy and signal indication graphs for different movements. Such plots provide system operators with the information needed to assess the efficiency of phase operations and to continuously monitor the level of green time utilization for different phases. Two applications to demonstrate how the data logging device can be used to monitor intersection operations are presented in this report. The first application is microscopic time-occupancy and signal indication status plots for different movements. Such plots provide system operators with the information needed to assess the efficiency of phase operations and to continuously monitor the level of green time utilization for each phase. The research project examined the validity of using this microscopic detector occupancy and signal indication status data to obtain traffic counts, identify heavy vehicles in the traffic stream, and determine the percentage of stopped and non-stopped vehicles. The second application is macroscopic in nature and is intended to show how the data logging device can be used to estimate average values of different performance measures based on detector and signal indication status information. The delay and speed results estimated using the proposed approach are compared to speed and delay data obtained from a VISSIM microscopic simulation model. The comparisons show that the data logger device can reliably and accurately estimate average delay and speed values for signalized intersection approaches using detector occupancy and signal indication data. 17. Key Words Traffic data, traffic signal controllers, signalized intersections 18. Distribution Statement Unrestricted; Document is available to the public through the National Technical Information Service; Springfield, VT. 19. Security Classif. (of this report) Unclassified Form DOT F 1700.7 (8-72) 20. Security Classif. (of this page) Unclassified Reproduction of completed page authorized 21. No. of Pages 51 22. Price

TABLE OF CONTENTS TABLE OF CONTENTS... I LIST OF TABLES... II LIST OF FIGURES... II 1. INTRODUCTION... 1 1.1 BACKGROUND... 1 1.2 PROJECT OVERVIEW AND OBJECTIVES... 2 1.3 REPORT ORGANIZATION... 3 2. REAL-TIME MONITORING OF CONTROLLER OPERATION... 4 2.1 OVERVIEW... 4 2.2 DATA LOGGING DEVICE: COMPONENTS AND COMMUNICATION ARCHITECTURE... 4 2.3 TRAFFIC CABINETS: AN OVERVIEW... 5 2.4 CONNECTING THE DATA LOGGING DEVICE TO DIFFERENT CABINET ASSEMBLIES... 8 2.5 DATA LOGGING DEVICE OUTPUT FILES... 12 2.6 MOES REPORTING CAPABILITIES OF TRAFFIC CONTROL SOFTWARE: STATE OF THE PRACTICE... 14 3. ESTIMATING DELAY AND SPEED USING DETECTOR DATA... 16 3.1 DEFINITIONS... 16 3.2 MOE ESTIMATION APPROACH... 17 3.2.1 Estimating Average Delay using Detector Data... 18 3.2.2 Estimating Average Speed using Detector Data... 20 4. RESEARCH METHODOLOGY AND EXPERIMENTAL DESIGN... 22 4.1 PROPOSED DELAY ESTIMATION METHOD... 22 4.2 PROPOSED SPEED ESTIMATION APPROACH... 25 4.3 HARDWARE-IN-THE-LOOP SIMULATION MODEL... 27 5. ANALYSIS AND RESULTS... 30 5.1 INTRODUCTION... 30 5.2 MICROSCOPIC TIME-OCCUPANCY AND SIGNAL INDICATION PLOTS... 30 5.2.1 Estimation of Vehicle Count and Vehicle Type... 33 5.2.2 Estimation of Stopped and Non-Stopped Vehicles... 38 5.3 DELAY AND SPEED ESTIMATION... 39 An Intersection Data Collection Device Utilizing Logging Capabilities... i

5.3.1 Delay Estimation Webster Formulation (Method1)... 39 5.3.2 Delay Estimation Method 2... 42 5. 3.3 Speed Estimation... 44 5.4 SUMMARY... 46 6. CONCLUSIONS AND FURTHER RESEARCH... 47 6.1 SUMMARY... 47 6.2 CONCLUSIONS... 48 6.3 FURTHER RESEARCH... 48 7. REFERENCES... 50 LIST OF TABLES TABLE 1: MOES REPORTING CAPABILITIES OF DIFFERENT CONTROL SOFTWARE PACKAGES... 15 LIST OF FIGURES FIGURE 1 DATA LOGGING DEVICE COMPONENTS... 5 FIGURE 2 NEMA TS2 TYPE 1 TRAFFIC CONTROL CABINET... 6 FIGURE 3 NEMA TS2 TYPE 2 TRAFFIC CONTROL CABINET... 7 FIGURE 4 PROPOSED DATA LOGGING DEVICE CONNECTION TO NEMA TS1 CABINETS... 9 FIGURE 5 TWO PROPOSED DATA LOGGING DEVICE CONNECTION OPTIONS TO TS2 TYPE 1 CABINET... 10 FIGURE 6 DATA LOGGING DEVICE PROPOSED CONNECTION TO TS2 TYPE 2 CABINET...11 FIGURE 7 SAMPLE OF DATA LOGGING DEVICE OUTPUT FILES... 13 FIGURE 8 DELAY COMPONENTS AT A SIGNALIZED INTERSECTION APPROACH... 17 FIGURE 9 TIME DISTANCE DIAGRAM AT A SINGLE SIGNAL (SKABARDONIS ET AL., 2005).. 24 FIGURE 10 ASSUMED FLOW DENSITY DIAGRAM (SKABARDONIS ET AL., 2005)... 25 FIGURE 11 HARDWARE-IN-THE-LOOP SIMULATION MODEL... 28 FIGURE 12 VISSIM SIMULATION NETWORK... 29 FIGURE 13 EXAMPLES OF TIME-OCCUPANCY PLOTS FOR TWO CONFLICTING PHASES... 31 FIGURE 14 ESTIMATED BOUNDARY LINES OF TIME-OCCUPANCY BETWEEN CARS AND HVS (1 ST VEHICLE IN THE QUEUE)... 34 FIGURE 15 ESTIMATED BOUNDARY LINES OF TIME-OCCUPANCY BETWEEN CARS AND HVS An Intersection Data Collection Device Utilizing Logging Capabilities... ii

(2 ND VEHICLE IN THE QUEUE)... 35 FIGURE 16 ESTIMATED BOUNDARY LINES OF TIME-OCCUPANCY BETWEEN CARS AND HVS (3 RD VEHICLE IN THE QUEUE)... 35 FIGURE 17 ESTIMATED BOUNDARY LINES OF TIME-OCCUPANCY BETWEEN CARS AND HVS (4 TH VEHICLE IN THE QUEUE)... 36 FIGURE 18 ESTIMATED BOUNDARY LINES OF TIME-OCCUPANCY BETWEEN CARS AND HVS (5 TH VEHICLE IN THE QUEUE)... 36 FIGURE 19 ESTIMATED BOUNDARY LINES OF TIME-OCCUPANCY BETWEEN CARS AND HVS (6 TH VEHICLE IN THE QUEUE)... 37 FIGURE 20 ESTIMATED BOUNDARY VALUES OF OCCUPANCY TIME FOR CARS AND HVS.. 37 FIGURE 21 A SAMPLE OF DISCHARGE TIME-HEADWAY IN A CYCLE... 39 FIGURE 22 COMPARISON OF SIMULATED AND ESTIMATED DELAY (METHOD 1)... 40 FIGURE 23 MEAN ABSOLUTE ERROR AND MEAN ABSOLUTE PERCENT ERROR - DELAY ESTIMATION (METHOD 1)... 41 FIGURE 24 COMPARISON OF SIMULATED AND ESTIMATED DELAY ESTIMATION (METHOD 2)... 42 FIGURE 25 MEAN ABSOLUTE ERROR AND MEAN PERCENT ABSOLUTE ERROR - DELAY ESTIMATION (METHOD 2)... 43 FIGURE 26 COMPARISON OF SIMULATED AND ESTIMATED SPEED ESTIMATION... 44 FIGURE 27 MEAN ABSOLUTE ERROR AND MEAN PERCENT ABSOLUTE ERROR OF SPEED ESTIMATION... 45 An Intersection Data Collection Device Utilizing Logging Capabilities... iii

1. INTRODUCTION 1.1 Background Delay and speed are the primary measures of effectiveness (MOEs) used to evaluate the performance of traffic signal systems. Delay and speed values, measured in the field, are extremely valuable to system operators because they provide accurate information regarding the quality of service for different movements at signalized intersections. Several methods have been developed and employed to measure delay in the field. However, most of these methods rely on manually collected traffic counts and require intensive data collection efforts. Detector and signal indication information, available in the traffic controller and cabinet infrastructure, can be effectively used to estimate delay and speed values as well as other signalized intersections MOEs. Real-time monitoring of traffic signal system operations can also be accomplished through the central or closed loop software that communicates with traffic controllers in the field through a network of communication devices. These control software tools use detector and signal status data to estimate different performance measures such as detector occupancy, volume, delay, speed, and the level of green-time utilization for each movement and for the intersection. There are a few significant issues with this approach. First, most of these control software tools report only average values over a specific time interval that ranges from one minute to fifteen minutes. This is a huge limitation when one wants to evaluate second-by-second performance of either the control logic or detection technology used. Second, the only data available are those explicitly collected by the vendor of the closed loop or central control software. In addition, these data are typically not easily accessible and may require complicated direct database access within the software. Finally, the type and accuracy of the measures obtained are highly dependent on the detection configuration used in the intersection. An Intersection Data Collection Device Utilizing Logging Capabilities... 1

1.2 Project Overview and Objectives This research project presents an alternative approach to achieve real-time monitoring of signalized intersections operations using instrumentation at each signalized intersection cabinet. A logging device is used to collect high-resolution detector and signal status data. The data logging device can be embedded in the cabinets and connected to the input and output (I/O) communication channels. The device logs the status of each I/O channel for every time interval, which can be as low as ten milliseconds (0.01 second), and stores it into a data file. This data file can be remotely accessed through an Internet Protocol (IP) based communication. This device allows for real-time high-resolution data collection and monitoring of signalized intersections operations independent of the control software used and, thus, has several potential advantages. The data items that can be monitored and reported are not limited by what data items are collected or by the frequency at which the vendor of the closed loop or central system software collects them. The interface device can be accessed from the district office over any communication channel available in the field. The proposed data collection device has several potential advantages: 1. The data items that can be monitored are not limited by what data items are polled by or the frequency they are polled, by the vendor of the closed loop or central system software, if any. 2. Intelligent data acquisition devices can be embedded in the signal cabinet that execute data tabulation logic and are accessible via web browsers over an IP based communication. 3. The system will be completely isolated from the ITD operations and will not impact the operation of the ITD signal systems. Two applications to demonstrate how the data logging device can be used to monitor intersection operations are presented in this report. The first application is microscopic time-occupancy and signal indication status plots for different movements. Such plots provide system operators with the information needed to assess the efficiency of phase operations and to continuously monitor the level of green time utilization for each phase. The research project examined the validity of using this microscopic detector occupancy An Intersection Data Collection Device Utilizing Logging Capabilities... 2

and signal indication status data to obtain traffic counts, identify heavy vehicles in the traffic stream, and determine the percentage of stopped and non-stopped vehicles. The second application is macroscopic in nature and is intended to show how the data logging device can be used to estimate average values of different performance measures based on detector and signal indication status information. The delay and speed results estimated using the proposed approach are compared to speed and delay data obtained from a VISSIM microscopic simulation model. The comparisons show that the data logger device can reliably and accurately estimate average delay and speed values for signalized intersection approaches using detector occupancy and signal indication data. This project has the followings four objectives: 1) review and document the state of the practice of real-time monitoring of traffic signal system operations; 2) test the validity of using the data logging device to monitor and report the status of different I/O channels; 3) develop a procedure to use detector occupancy and signal indication data, reported by the data logging device, to estimate approach and intersection performance measures; and 4) validate the procedures to estimate performance measures and test their accuracy using a hardware-in-the-loop simulation model. 1.3 Report Organization This report is organized in six chapters. After the introduction, chapter 2 presents a background on real-time monitoring of controller operation and the state of the practice in traffic signal systems real-time monitoring. Chapter 3 includes a review of different methods used to estimate speed and delay based on detector data. Chapter 4 covers the analysis methodology and experimental design. Chapter 5 documents the results of the analysis. Finally, chapter 6 presents the conclusions and proposed ideas for future research. An Intersection Data Collection Device Utilizing Logging Capabilities... 3

2. REAL-TIME MONITORING OF CONTROLLER OPERATION 2.1 Overview In integrated traffic signal systems, real-time monitoring can be accomplished through the central or closed loop software that provides control decisions and continuously communicates with the traffic controllers and cabinets. The type and quality of the MOEs reported depend on the control software and on the configuration of the detection system used in the field. The only MOEs available are those explicitly collected by the vendor of the closed loop or central system. In addition, only average values are usually collected and reported. Furthermore, data retrieved by the closed loop or central system are not typically accessible to users. For traffic signal systems that have no control software, real-time monitoring can only be done by accessing the MOEs collected and stored in the traffic controller. System operators have to access these controllers and manually download this data every time. This is a huge limitation considering the limited resources available for operators of such small traffic signal systems. An alternative to achieve real-time monitoring is to use instrumentation at each cabinet that is connected to actuator and detector signals. The intelligent data acquisition device provides real-time high-resolution data logging and performance monitoring for signalized intersections. It can be embedded in the signal cabinet and executes data tabulation logic and writes the status of all I/O channels to a data file that is remotely accessible through IP based communication. The data that can be monitored are not limited by what data are collected or by the frequency at which they are collected by the closed loop, central system software, or traffic controllers. 2.2 Data Logging Device: Components and Communication Architecture Figure 1 shows the proposed data logging instrumentation and its major components. This instrumentation is based upon the Opto 22 family of ultimate I/O brains (item 10 in Figure 1) and SNAP IDC 5 modules (4 channel10-24 VDC inputs- item 3 in Figure 1). An Intersection Data Collection Device Utilizing Logging Capabilities... 4

Figure 1: Data logging device components. 2.3 Traffic Cabinets: An Overview A traffic cabinet is essentially a platform within which modular components can be added to serve a variety of applications at the intersection. It provides the communications infrastructure between the various subsystems, as well as a system to monitor their operation. Further, the cabinet provides power supplies suitable for the various electronic subassemblies mounted throughout the cabinet. Cabinet assemblies consist of a controller cabinet, controller unit, back panel, malfunction management unit, bus interface unit, switches, and connectors. The National Electrical Manufacturers Association (NEMA) family of cabinets include: NEMA TS 1, NEMA TS2 Type 1, and NEMA TS2 Type 2 cabinets. NEMA TS1 cabinets include a controller along with the conflict monitor, detectors connection matrix, load switches, other peripheral equipment, and the necessary internal wiring. NEMA TS2 standard defines two types of controllers and cabinet architectures, the TS2 Type 1 and TS2 Type 2. The NEMA TS2 controller assembly is nearly identical to the TS 1. The two primary An Intersection Data Collection Device Utilizing Logging Capabilities... 5

differences are the change in controller unit and the conflict monitor being replaced by a malfunction management unit (MMU). The NEMA TS2 Type 1 cabinet is unique in the sense that it uses a RS-485/SDLC data link connection to the peripheral devices, with a separate power connector. The TS2 Type 2 provides the same connectors as the TS1 but includes the data link connector. The TS2 cabinet also uses a bus interface unit (BIU) for communication between the various control components and detectors. The BIU provides simplification in cabinet wiring as well as flexibility and power. The TS2 assembly contains a shelf-mounted power supply unit that provides the appropriate power to each of the controller devices. The detectors in the TS2 cabinet are rack-mounted. The TS2 standard defines advanced traffic signal operations, such as coordination and preemption, and developed standards for pre-timed operations and advanced cabinet monitoring and diagnostics. Details of NEMA TS2 Type 1 and NEMA TS2 Type 2 cabinets are shown in Figure 2 and Figure 3, respectively. Figure 2: NEMA TS2 Type 1 traffic control cabinet. An Intersection Data Collection Device Utilizing Logging Capabilities... 6

Figure 3: NEMA TS2 Type 2 traffic control cabinet. In terms of communication with the traffic controllers, the NEMA TS1 and NEMA TS2 Type 2 standards use the four connectors A, B, C and D on the front of the controller. The A connector provides power to the controller as well as inputs and outputs with the cabinet. The B and C connectors provide various inputs and outputs for control. The A, B, and C connector pin outs are standardized by NEMA and are interchangeable among all manufacturers. Each connector is different, preventing cables from being inserted in the wrong connection port. The D connector provides communication, preemption, and expanded detection capabilities that are used in more advanced systems. Typical controllers have eight available detection inputs. The D connector provides input for eight additional detectors. The D connector pin out is not standardized by NEMA; therefore, it may not be interchangeable. In NEMA TS2 cabinets, the BIU links the controller to the cabinet input/output (I/O) elements. It can also be used as a detector interface device. The BIU is responsible for controlling load switches, receiving and isolating pedestrian calls, analyzing detector faults, time-stamping detector calls, and providing detector resets. By design, the BIU is An Intersection Data Collection Device Utilizing Logging Capabilities... 7

free of operator controls. The BIU performs its I/O functions based upon a pre-wired card rack address. The MMU is a more advanced device, not only monitoring all of the conflict voltages, but also communicating with the controller, providing an additional element of monitoring. The type-16 MMU is usually used in a NEMA TS2 standard cabinet that monitors up to 16 traffic signal channels for conflicting inputs, improper sequencing, incorrect timing, and invalid signal voltage levels. The MMU is also capable of operating in older TS1 type cabinets and is compatible with 12-channel conflict monitor units conforming to the TS 1 standard. All connectors, indicators, and operator controls are located on the front panel of the MMU. Channel and control input signals and relay output connections are made through two connectors. Indicators on the front of the MMU provide status and fault information. The MMU performs continuous diagnostic tests during all operating modes. 2.4 Connecting the Data Logging Device to Different Cabinet Assemblies In a standard NEMA TS1 style cabinet, the connections to the controller are made through the connection matrix on the back panel of the cabinet. These terminals are the only available connection points for the data logging device. The proposed connection is shown in Figure 4. The data logging device cables should have non-locking fork terminals that can be connected to the matrix. The connection is done by loosening the screws on the back panel then connecting the data logging device cable terminals. This should not interfere with the cabinet operations and should not cause any malfunction within the cabinet. An Intersection Data Collection Device Utilizing Logging Capabilities... 8

DLD CMU Controller A B C D Load Switch Detector Auxiliary Devices MS Connector Connection Matrix Figure 4: Proposed data logging device connection to NEMA TS1 cabinets. In a NEMA TS2 Type 1 style cabinet, the data logging device connection is rather challenging as the cabinet assembly does not have a connection matrix. The controller communicates with the cabinet using a serial connection through the cabinet s BIUs. The communication link from the controller uses the RS-485 serial communication format or synchronous data link control (SDLC) in combination with the NEMA standard TS2 command frames. There are two possible connection options. The first option is to connect via the data logging device to the terminals on the back panel of the cabinet. The second option is to connect the cabinets BIUs that are hardwired to this back panel. This will likely require cooperation with the cabinet vendors as details of BIU wiring mechanism are needed. The first mode of connection is represented uses a solid line and the second mode is represented uses dashed lines in Figure 5. An Intersection Data Collection Device Utilizing Logging Capabilities... 9

SDLC DLD MMU Controller BIU Load Switch BIU Detector BIU Auxiliary Devices Back Panel SDLC Figure 5: Two proposed data logging device connection options to TS2 Type 1 cabinet. As shown in Figure 6a and 6b, there are two options to connect the data logging device in a NEMA TS2 Type 2 style cabinet since this cabinet combines standards for the TS1 and TS2 Type 1. The first is to connect the device to the connection matrix on the back panel, like that of the NEMA TS1 cabinet. The second is through a serial connection through either the cabinet s back panel or the BIUs similar to that for TS2 Type1 cabinets. An Intersection Data Collection Device Utilizing Logging Capabilities... 10

New SDLC DLD MMU Controller BIU Load Switch BIU Detector BIU Auxiliary Devices Connection Matrix MS Connector Back Panel SDLC a. Connection Via the Back Panel and/or BIUs DLD MMU Controller Load Switch BIU Detector BIU Auxiliary Devices MS Connector Connection Matrix b. Connection Via Connection Matrix Figure 6: Data logging device proposed connection to TS2 Type 2 cabinet. An Intersection Data Collection Device Utilizing Logging Capabilities... 11

2.5 Data Logging Device Output Files The data logging device monitors and records the communication exchanged between the detector and the controller and between the controller and signal heads. It also records any other special calls sent to the controller such as pre-emption calls. In essence, the device monitors activities in all input and output communication channels to and from the controllers. In each sampling interval, it scans the status of all input/output channels and records the state of each channel (on or off). The data are then stored in a log file which can be accessed through the Ethernet port. The sampling interval for data logging can be as small as 10ms. However, since most cabinets update the communication channel status every 300 ms, a resolution time ranging from 300 ms to 1000 ms can more easily be used in traffic signal system monitoring applications. Data recorded by the data logging device include date, time, and the status of each communication channel on the sampling interval. Figure 7 shows a sample of the data logging device files for the status of detector and signal indication I/O communication channels. A value of -1 represents when the communication channel is On ; a value of 0 represents when the communication channel is Off. Figure 7a shows the status of different vehicle detectors using a 100 ms resolution. Detector occupancy and vehicle count can be directly calculated from these raw detector data principally based on the discontinuity distribution of occupancy time followed by un-occupancy time. Figure 7b shows the signal indication status for different phases. The average cycle length and the duration of green, red, and yellow intervals can be directly calculated from the raw signal state and timing data. An Intersection Data Collection Device Utilizing Logging Capabilities... 12

a. Status of Detector Input Channels b. Status of Signal Indication Output Channels Figure 7: Sample of data logging device output files. An Intersection Data Collection Device Utilizing Logging Capabilities... 13

2.6 MOEs Reporting Capabilities of Traffic Control Software: State of the Practice Performance measures reporting capabilities for different controller and control software tools were reviewed and documented in this section. Two controller software tools: Econolite (used in Econolite TS2 controllers) and Nextphase (used in 170 and 2070 type controllers) and two centralized control software packages QuicNet/4 and Icons TM are reviewed. Their MOEs reporting capabilities are listed in Table 1. An Intersection Data Collection Device Utilizing Logging Capabilities... 14

Table 1: MOEs Reporting Capabilities of Different Control Software Packages MOEs Reporting Volume Delay Speed Occupancy Controller Software Centralized control Software Econolite Nextphase QuicNet Icons TM When system detector is enabled, volume is reported in the detector events logs. Speed are calculated based on average vehicle and detector length (single speed detector); or based on effective distance between the leading edges of start and end detectors (speed trap length) and the time (used in two-detector). When system detector is enabled, occupancy is reported in the detector events logs. Reports actual counts of the detector during the most recent reporting period for all detectors Reports average speed for system detectors only. Speed samples are registered at the end of each actuation. It is shown in system detector status of submenu of Status Reports the percentage time each detector was occupied during the most recent reporting period for all detectors. Reports volume counts for system detectors only. Average delay based on detector occupancy Reports average speed for system detectors only. Reports average occupancy for system detectors only. Reports actual counts of the detector during the most recent reporting period for all detectors. Reports volume for each link Average delay for each link based on detector occupancy Calculated speed value using a measured volume and occupancy in a specific time period, it depends on detection zone length and vehicle length. Reports average link speed. Reports detector occupancy during the most recent reporting period for all detectors. Reports volume for each link Green Split Actual green split is reported. It is calculated via phase split minuses clearance time. Reports minimum green split, nominal green split and maximum green split for each phase. Provides real time split monitoring Reports real-time green split display for each link Time-Space Diagram Displays a real-time space diagram. Shows green, yellow and red times and progression of vehicle An Intersection Data Collection Device Utilizing Logging Capabilities... 15

3. ESTIMATING DELAY AND SPEED USING DETECTOR DATA 3.1 Definitions Control delay at a signalized intersection approach is defined by the Highway Capacity Manual (HCM2000) as the additional travel time experienced by a vehicle affected by intersection control, relative to conditions where the vehicle is unaffected by intersection control. The following definitions of delay and speed are used in this project (Figure 8). Control delay ( d t ) is the portion of the total delay attributed to traffic signal operation for signalized intersections. Control delay includes four components: 1) initial deceleration delay ( d d ), 2) queue move-up time, 3) stopped delay ( d ), and 4) acceleration delay ( d ). a s Approach delay ( d ap ) includes stopped time, but also includes the time lost when a vehicle decelerates from its original speed to a stop, as well as accelerating from the stop back to its original speed. Stopped delay ( d ) is the time that a vehicle is stopped while waiting to pass s through the intersection. It includes only the time that a vehicle is actually stopped waiting at the red signal. Deceleration delay ( d ) is defined as the time needed by a vehicle to reduce its speed. d Acceleration delay ( d ) is defined as the time taken by a vehicle to resume its a desired speed from a stop. This delay can begin before or at the stop bar, depending on the vehicle s queue position. The acceleration delay consists of two components: acceleration before the stop bar ( d a1) and acceleration after the stop bar ( d a2 ). An Intersection Data Collection Device Utilizing Logging Capabilities... 16

Queue delay ( d q ) is defined in HCM as the delay experienced by queued vehicles. This delay consists of two control delay components d s and d a1. Figure 8: Delay components at a signalized intersection approach. The following speed definitions are used in this study: Average Speed(s) is the summation of the instantaneous or spot-measured speeds at a specific location of vehicles divided by the number of vehicles observed. Average Running Speed (s r ) is defined as the length of the segment divided by the average running time of vehicles to traverse the segment. "Running time" includes only time that vehicles spend in motion. Average Travel Speed (s t ) is defined as the length of the segment divided by the average travel time of vehicles traversing the segment, including all stopped delay times. 3.2 MOE Estimation Approach A variety of methods to estimate different MOEs for an intersection approach have been developed using data collected through loop or video detectors. The following sections An Intersection Data Collection Device Utilizing Logging Capabilities... 17

document several delay and speed estimation methods using detector and signal indication data. 3.2.1 Estimating Average Delay using Detector Data Skabardonis et al. (2005) proposed an analytical model to estimate the total control delay of an intersection approach in real-time based on flow, occupancy measurements, and signal status data. The model is based on the kinematic wave theory that considers the temporal and spatial formation of the queue and the assumption of a linear flow-density relationship. This delay was considered as the sum of 1) the delay because of a traffic signal, 2) the delay because of the queue, and 3) the over-saturation delay. The detectors were placed approximately 300 feet upstream of the intersection stop-line, and detector data were collected and stored every 30 seconds. The model was applied in two arterial sites, and the predicted results were compared with the simulated data from COSSIM and the field data. This project uses this method and it will be described in detail in chapter 4. Liu et al. (2005) proposed a method to estimate average stopped delay of an intersection approach using flow measurements and arrival timings from the two loop detectors at the beginning and end points of the approach segment. Hellinga et al. (2000) proposed a regression-based approach to estimate the total control delay of an intersection approach using occupancy data from detectors located at different distances relative to the approach s stopbar. Loop detectors were modeled at four different locations (5, 30, 100 and 250 m upstream from the stopbar), and three different data aggregation intervals (100 seconds, 300 seconds, 900 seconds) were considered. Simulation models were used to generate data needed to calibrate the regression model parameters. These regression models were then used to estimate the average delay under different traffic volume conditions based on detector occupancy data. One major limitation of this method is that it did not consider signal timing parameters, such as average cycle length and green time to cycle length ratio, which could greatly influence delay. Li et al. (2008) proposed a formulation for average control delay estimation by cycle for signalized intersections. The delay is expressed as a function of saturation flow rate, start of green indication, lost time, duration of green interval for each cycle, queue clearance An Intersection Data Collection Device Utilizing Logging Capabilities... 18

time, arrival count, and free flow travel time from the advance loop to the stopbar, as shown in Equation (1). d j [ t 2 j CQ ( g j T Lost )][ t j CQ ( g j T Lost j c ) j 1 t c 1 2( c A( t T j 1 FF ) 1)] t j tcq j 1 c 1 A( t T FF ) ( t c j 1 ) (1) where d j = the average control delay for cycle j; µ t j CQ = the clearance queue time for cycle j; g j = the green start time for cycle j; T Lost = the lost time; c j = the end time for cycle j; c j-1 = the end time for cycle j-1; A(t) =the instant arrival count at advance loops at time t; T FF = the free flow travel time from advance loop to the stop line. Kebab et. al (2007) proposed to estimate the total control delay by collecting individual vehicle s timestamps at three locations along the intersection approach. The three data collection points are: 1) at a point beyond the maximum queue length, 2) at a point where the turning movements are fully developed, and 3) at the approach s stopbar. The delay is treated as the sum of the differences between the actual and free flow travel time at two segments among the three points. Tung (2007) applied this method to estimate field delay using video detection. The results of his study showed that the automated delay measurement procedure produces accurate and reliable delay estimates. When compared An Intersection Data Collection Device Utilizing Logging Capabilities... 19

against delay values measured using the procedure proposed in the Highway Capacity Manual, the proposed automated delay measurements produced more accurate results. Lin et al. (2004) proposed an approach to estimate the control delay at signalized intersections. The approach reduces the delay at each intersection, a non-negative continuous variable, into two distinctive states, a state of zero-delay and a state of nominal delay, coupled with a one-step probability transition matrix that relates the delay to a vehicle to its delay at the adjacent upstream intersection. The calibration of the parameters in the one-step probability transition matrix is based on the flow level, the flow composition, and the degree of signal coordination along the path of a trip. 3.2.2 Estimating Average Speed using Detector Data Son et al. (1998) classified the vehicles in a cycle into two categories according to discharge headways: vehicles in the queue with saturation flow headways and vehicles after the dissipation of the queue with departure headways equal to arrival headways. The average speeds for the two types of vehicles can be calculated principally using detector occupancy and vehicle count data. Finally, on the basis of the two speeds and signal timing, the average speed for each cycle is estimated. This method is used as the speed estimation method in this project and will be described in detail in chapter 4. Zhang (1999) proposed a model to estimate the average speed for arterial traffic by combining two speeds: one is estimated based on the approach s volume/capacity ratio and the other is based on volume counts and detector occupancy data. Weighting factors are chosen to combine the two speeds. The average speed for each approach is a weighted average of the two speeds. Weighting factors are determined and calibrated based on field measurements of speed and according to the traffic volume level on the approach. Wang et al. (2000) proposed a simplified equation to estimate arterial speed by isolating the effect of speed variance. They conducted a study on the speed variance for different volume levels and found that the variance is inversely proportional to volume levels. They also found high correlation between speed variance and the mean effective vehicle length. They established a log regression model to improve the accuracy of speed estimation based on detectors occupancy data. Zhang et al. (2006) applied the catastrophe An Intersection Data Collection Device Utilizing Logging Capabilities... 20

theory to estimate average speed for the relationship among the traffic variables associated with speed, occupancy, and flow. These variables could be extrapolated from the data obtained from a single loop detector using three simple linear transformations. Bermejo et al. (2003) used the extended Kalman filter method to linearize the measurement equation of a general Kalman Filter model for estimating average speed based on detector data. There are two phases in this method: the time update phase is operated to predict a new state, and the measurement update phase is operated to correct any new state. Lucas et al. (2004) proposed an approach to estimate average speed on arterial based on second-by-second data from upstream and downstream detectors. The detector data are first used to identify platoons of vehicles and then a matching algorithm compares the platoons identified at the upstream and downstream locations. The average speed estimate is based on the travel time of the median vehicles in the platoons as determined at both the upstream and downstream locations. Sun et al. (1999) proposed a model to estimate average speed using single loop inductive waveforms. This model uses signal processing and statistical methods to extract speeds and involves two main procedures. The first is the extraction of the vehicle slew rate from the inductive vehicle waveform signal from the detector. The second is the estimation of the vehicle speed based on slew rate of each vehicle. While this method yielded high accurate results, it requires special instrumentation for each detector and is highly sensitive to the accuracy of the detector s inductive signal. An Intersection Data Collection Device Utilizing Logging Capabilities... 21

4. RESEARCH METHODOLOGY AND EXPERIMENTAL DESIGN 4.1 Proposed Delay Estimation Method Two different delay estimation methods are used in this project. The first is based on the Webster delay equation (Liu, et al., 2005), a commonly used model to determine the average control delay of an intersection approach. The Webster delay equation has three terms. The first term presents the average delay for a particular approach assuming uniform arrivals at a fixed-time signal-controlled intersection and can be easily derived using deterministic queuing theory. The second term is added to account for random arrivals. The third term is subtracted from the first two terms and varies from zero to a value equal to the second term. The Webster delay equation is given as follows: C(1 2(1 ) ) x 2q(1 2 2 1/ 2 (2 5 ) D 0.65( ) x (2) 2 x) C q where D = Average control delay (seconds/vehicle); C = signal cycle length (seconds); x = degree of saturation; q = volume (vps); and λ= effective green proportion. Parameters C and λ are obtained from the signal timing data. Volume, q, is directly obtained from the detector measurement. The degree of saturation, x, should be calculated based on detector occupancy and signal timing data. A stop-line detector is required to collect flow and occupancy measurement. The second delay estimation method examined in this study uses the analytical model proposed by Skabardonis and Geroliminis (2005). In this model, the delay is the sum of An Intersection Data Collection Device Utilizing Logging Capabilities... 22

two types of delay: delay caused by the traffic signal and delay caused by the queue present in the intersection approach. The first part of the model assumes that each vehicle has no interaction with other vehicles in the traffic stream. Under this assumption, all queued vehicles are considered stopped at the stop line (vertical queue). The delay, (d (t)), of a single vehicle as a function of arriving time, t, is given by Equation (3): u f u f d( t) r T t (3) 2 2 d a where r = the effective red time; T = the driver s reaction time; u f = free flow speed; γ d = the vehicle s deceleration rate; γ a = the vehicle s acceleration rate; and t = the time a vehicle starts to decelerate. The parameters for γ d, γ a, u f and T are assumed constant. Their values are determined according to the Institute of Transportation Engineers (ITE) guidelines. The effective red time, r, is directly obtained from signal timing data reported in the data logging device output files. The parameters for u f and t are obtained from detector measurements. The delay in the second part of the Skabardonis and Geroliminis model is the result of the queue present at the traffic signal approach. It is estimated based on the kinematic wave theory considering the temporal and spatial formation of the queue and assuming a relationship between linear flow and density. The queue delay, d q, of the n-th vehicle arriving at the signal from the beginning of the red time is the sum of three types of delays (d q1, d q2, d q3 ) as illustrated in Figure 9. An Intersection Data Collection Device Utilizing Logging Capabilities... 23

Figure 9: Time distance diagram at a single signal (Skabardonis et al., 2005). The delay values can be determined using the following equations: L d (min( n, N ) 1) s q1 qm (4) u f L d ) s q2 (min (max ( n, Nqm), Nq) Nqm (5) uw Ls d q 3 (min( n, N q ) 1) (6) w where L s = the effective length of a stopped vehicle; u w = speed of the shockwave; w = congested wave speed; N qm = number of the maximum queue; and N q = number of the maximum back of the queue. An Intersection Data Collection Device Utilizing Logging Capabilities... 24

The effective length of a stopped vehicle, L s, is assumed constant using the reciprocal of the jam density (k j ). The parameters, u w and w, are obtained from flow and occupancy measurements based on an assumed linear flow-density relationship shown in Figure 10. Figure 10: Assumed flow density diagram (Skabardonis et al., 2005). The parameters of N qm and N q can be calculated using Equations (7) and (8). L qm r u u f f u w u w L qm N qm (7) Ls L q r w wu w u w L q N q (8) Ls 4.2 Proposed Speed Estimation Approach The average approach speed is estimated using the model proposed by Son and Oh (1998). Vehicles in each cycle are classified into two categories: vehicles in the queue with saturation flow headways and vehicles arriving and departing after the dissipation of the queue with arrival headways. The average speed of a cycle (V cycle ) is determined using the following equation: An Intersection Data Collection Device Utilizing Logging Capabilities... 25

V cycle R l t 0 v NS N ( N S s N 1) A v NA N A (9) where t 0 = the time of the first vehicle s arrival during red time; R = the red time; v NS = the average speed of vehicles discharging at the saturation flow rate; v NA = the average speed of vehicles discharging at the arrival flow rate; N S = the number of vehicles crossing the stop line with saturation headway in a cycle; N A = the number of vehicles crossing the stop line with arrival headway in a cycle; l = the average length of the sum of vehicles and detectors. Red time, R, is directly obtained from signal timing data reported in the data logging device output file. The parameters, t 0, N S, N A and N C, are obtained from detector data. The parameter, v NA, can be calculated using the following equations: v i ( t l occ ) i (10) v NA N i A 1 N v A i (11) v NS is calculated using the equation, An Intersection Data Collection Device Utilizing Logging Capabilities... 26

v NS N ( Tocc) cycle ( R t0 s l ) l v NA N A (12) In the three equations, V i = the speed of vehicle i; (T occ ) cycle = the total occupancy time for a cycle; and (t occ ) i = the occupancy time of vehicle i. The parameters, (T occ ) cycle and (t occ ) i, are obtained directly from detector data. 4.3 Hardware-in-the-Loop Simulation Model The data logging device was tested and validated in the lab using a hardware-in-the-loop simulation model, in which the control of the intersection in the simulation model was done by an actual traffic controller. A controller interface device (CID) was used to facilitate the information exchange between the microscopic simulation model and an actual traffic controller. Detector actuation information was sent from the simulation model to the controller. Signal status information was sent back from the controller to the simulation model. This data exchange was done in every simulation time step. In this experiment, VISSIM microscopic simulation was used along with a NEMA TS2 traffic controller. The data logging device was connected to the controller through an interface connected to the A, B, C, and D connectors in the traffic controller. The data flow in the hardware-in-the-loop simulation model used in the analysis is shown in Figure 11. The simulation time step was set to 100 ms (0.1 second). An Intersection Data Collection Device Utilizing Logging Capabilities... 27

Phase status Detector actuations Detector Detector actuations PC actuations Phase status CID Phase status Controller Record data DLD Figure 11: Hardware-in-the-loop simulation model. The intersection used in the analysis is an isolated intersection located in the city of Moscow, Idaho. The VISSIM simulation network is shown in Figure 12. This intersection was run using standard eight-phase NEMA operation. Data presented in this project focused on the eastbound approach through traffic (phase 4). The approach has two lanes with a stop bar and an advanced detector placed approximately 180 feet upstream of the intersection stop bar. Each simulation ran for a total simulation time of 20 minutes; data were collected for the last 15 minutes only. The average value of five runs was used for each case. The measure of effectiveness (MOE) chosen for this experiment was average delay and speed of the eastbound through movement. The objective of the hardware-in-the-loop simulation model experiment was to determine whether the average delay and speed for the approach can be estimated with an acceptable level of accuracy using the highresolution data logging device output data. An Intersection Data Collection Device Utilizing Logging Capabilities... 28

Figure 12: VISSIM simulation network. An Intersection Data Collection Device Utilizing Logging Capabilities... 29

5. ANALYSIS AND RESULTS 5.1 Introduction This chapter presents the results and analysis of two applications to demonstrate how the data logging device can be used to monitor intersection operations. The first application uses microscopic time occupancy plots for different detectors. Validity of using detector occupancy and signal indication date to estimate vehicle count, vehicle type, and the numbers of stopped and non-stopped vehicles is examined as part of this analysis. The second application is macroscopic and involves using the characteristics of detector occupancy, headway, and signal indications to estimate average delay and speed values using the methods identified in chapter 3. Delay and speed values reported by the VISSIM hardware-in-the-loop simulation model were assumed to be the true delay and speed values. They were compared against values estimated using the data logging device output files. Two measures - mean absolute error and mean absolute percent error, were used to compare the accuracy of the estimated measurements to true values. 5.2 Microscopic Time-Occupancy and Signal Indication Plots Figure 13 shows an example of continuous time-occupancy plots for two detectors located on the stop bar of an intersection approach. Signal indication for the approach is shown along the x-axis. The plots are updated at 300ms intervals, a typical rate for a standard cabinet to update detector and signal status information. These plots can provide system operators with useful information regarding the efficiency of the phase operations. Information such as average detector un-occupancy time and green time utilization can be obtained directly for these graphs. An Intersection Data Collection Device Utilizing Logging Capabilities... 30