Investigating the Impact of Skewed Pneumatic Traffic-Counting Tubes on Accuracy

Transcription

1 Investigating the Impact of Skewed Pneumatic Traffic-Counting Tubes on Accuracy By Benjamin Weible Marshall University Abstract: Pneumatic counting tubes are used in several countries around the world for the purpose of measuring traffic volumes, speeds, and vehicle classification. Studies claim high accuracy; close to 99% when installed correctly (1). It is crucial to maintain this level of accuracy to continue utilizing pneumatic counting tubes for their simplicity and reliability. A 48-hour study was conducted in West Virginia to compare the traffic volumes and vehicle classification accuracy between three sets of tubes in a single lane spaced closely together. One set of tubes was installed correctly at 0 perpendicular to the traffic lane. The other two sets were installed at 5 and 10 to simulate an incorrect installation, such that the tubes are triggered by individual wheels as opposed to axles. A video camera was installed to record each vehicle as it passed over the three sets of tubes and served as the baseline for computing error in the tube counts. The study showed that the angled tubes generated total errors between 11%-18%, while the correctly installed tubes generated less than 1% total error. Errors in vehicle classification were also severe, due to individual wheels triggering the tubes to count each wheel as an axle. This can cause the tubes to classify passenger vehicles with two axles as heavy vehicles with four or more axles. Furthermore, this can cause the tubes to combine vehicles, in turn lowering the vehicle count. This report compiles the results of this study and provides expectations of the data when tubes are installed as discussed. A report prepared for the Mid-Atlantic Transportation Sustainability University Transportation Center Charlottesville, VA July 2016

2 2 TABLE OF CONTENTS 1. Introduction Background 4 3. Literature Review Data Collection 6 5. Data Analysis Summary References..23

3 3 LIST OF FIGURES Figure 1: MetroCount Roadside Unit..4 Figure 2: Tube Installation Diagram (1).5 Figure 3: A and B Tube Connectors 5 Figure 4: Survey Location in Huntington, WV.7 Figure 5: Tube Installation..7 Figure 6: Design Plan for Tube Installation..8 Figure 7: Crosswalk as 0 Reference.9 Figure 8: Measurements Before Installation.9 Figure 9: Completed Installation.10 Figure 10: NuMetrics NC-200 Installed 11 Figure 11: CountPRO Software. 11 Figure 12: Ergodex DX-1 COUNTpad. 12 Figure 13: CountCam System Installed..13 Figure 14: MetroCount Traffic Executive Software 14 Figure 15: Inspector Gadget Tool Displaying a Vehicle with 5 Axles..14 Figure 16: Vehicle Counts by 15-Minute Interval Before Adding Video Count Figure 17: Vehicle Counts After Adding Video Count...15 Figure 18: 5 Axle Semi-Truck Individual Report Comparison..16 Figure 19: 5 Axle Semi-Truck Recorded at 10:14:14 AM. 16 Figure 20: FHWA Vehicle Classification Scheme (6) Figure 21: 5 Axle Semi-Truck Inspector Gadget for 0 Tubes...18 Figure 22: 5 Axle Semi-Truck Inspector Gadget for 5 Tubes...18 Figure 23: 5 Axle Semi-Truck Inspector Gadget for 10 Tubes Figure 24: Motorcycle Individual Report Comparison...19 Figure 25: Motorcycle Recorded at 10:52:40 AM...19 Figure 26: Counts by Class Between All Three Sets of Tubes...20 Figure 27: How to Achieve Perpendicularity...21 Figure 28: Remove Duplicates Tool in Excel...21 LIST OF EQUATIONS Equation 1.6 Equation 2.6 Equation 3.6

4 4 1. Introduction Accurate measurement of traffic volume is important for traffic road design, traffic signal timing, and research purposes. To collect traffic volume data, pneumatic road tubes are often preferred for their simplicity, low maintenance design, detailed data collection, and accuracy. Pneumatic road tubes installed in tandem are capable of measuring axle spacing, which facilitates classifying vehicles into 13 vehicle types in accordance with the FHWA Vehicle Classification Scheme. Vehicle classification provides valuable information about how specific roads are being used. While pneumatic tubes have been used for many years to collect this data and they are known to be very accurate, improper sensor installation can reduce that accuracy. Since installation of the tubes requires a technician to be in and adjacent to the flow of traffic, it can be difficult to achieve a perfect installation every time. One common sensor installation problem is for the tubes to be skewed (i.e., not perpendicular to the lane of traffic). This study will investigate the impact of sensor skew on the accuracy of the collected data. Specifically, there are two objectives of this study. 1. Determine the effects of skewed pneumatic tube counters on the accuracy of traffic volume measurement and vehicle classification. Tubes will be installed at 0, 5, and Determine if data collected with skewed tubes can be salvaged to avoid repeating a survey. 3. Find signs in the collected data that may suggest that the tubes were installed askew. 4. Determine a threshold angle at which point the data may become unusable. 2. Background The pneumatic tube modules used in this study are the MetroCount 5600 (Figure 1). Figure 1: MetroCount Roadside Unit The diagram in Figure 2: Tube Installation Diagram (1) shows the layout of a proper tube installation across a bidirectional road.

5 5 Figure 2: Tube Installation Diagram (1) Pneumatic tubes use changes in air pressure to trigger sensors as vehicles go by. The tubes are made of a strong rubber that can be stretched and resist the wear of several thousands of vehicles. One end of each tube is inserted into the data collection module, and the other end is either tied in a knot or plugged to prevent air from escaping. The tubes are anchored into the asphalt by webbing wires that grip the tubes, which are then nailed to the road. The use of two tubes enables the tube module to record vehicle speed and axle spacing for vehicle classification. The tubes can be left to collect traffic data continuously for up to 290 days for continuous data collection (3). The MetroCount module has two connectors which are labeled A and B (Figure 3), and tubes are installed such that the first tube to be triggered by a passing vehicle is connected to side A, and the second tube to B. This allows the module to determine a vehicle s speed and classification based on the time between triggers and number of triggers within a certain timeframe. Figure 3: A and B Tube Connectors

6 There are more traffic counting devices available, but few of them are capable of classifying vehicles or doing so accurately. Some of them are accurate, but not automated, which require a person to count by hand. These devices include the TDC Ultra by Jamar Technologies, and the CountCam system by Counting Cars. The NC-200 by NuMetrics is automated, but is known to be affected by weather and cannot detect stopped vehicles (4). The advantage of pneumatic tubes amongst these is that they are automated and are capable of recording more detailed data in their analysis software, even in harsh weather conditions. 3. Literature Review Accuracy of Pneumatic Road Tube Counters A study conducted by McGowen and Sanderson explains the types of error commonly generated by the tubes, and also provides a framework for calculating each type of error, which serve as the primary framework for the accuracy reported in this study (1). In some cases, the tubes will not detect a vehicle as it passes over the tubes. While in other cases, the tubes can sometimes detect more vehicles than are actually present. This results in an absolute error, such that the tubes are counting incorrectly, yet the average of positive and negative errors cancel each other, resulting in a total error as little as 1%. McGowen and Sanderson provide several equations for calculating these types of error in their report. Equations 1-3 below will be used to calculate error percentages in this report. Equation 1 6 Equation 2 Equation 3 MetroCount Classifier Testing A situation in 2002 regarding inaccurate classification semi-trucks called for a software update in the MetroCount pneumatic tube counters and testing the update. The testing was performed and reported by Pell, who claimed a perfect 100% accurate classification of 93 high speed trucks after the update when compared to a video count. The report includes a customer s testimony regarding the results who states, Admittedly the sample was not that large, but we are not used to devices that can perfectly classify even a sample of 93 trucks. And although Bob has not yet looked at files from the other devices, they cannot have performed any better (and we would be surprised if they had performed as well), (2). 4. Data Collection Location and Duration This study was conducted on the outermost lane on the right side of a unidirectional main corridor in Huntington, West Virginia (Figure 4 and Figure 5).

7 7 Figure 4: Survey Location in Huntington, WV Figure 5: Tube Installation Traffic data was collected continuously for 48 hours beginning Thursday (7/21/2016) at midnight until Friday (7/22/2016) at 11:59 PM using the following methods. Weather was sunny and warm on Thursday, and Friday was mostly sunny with a brief, 2-hour rain storm in the afternoon. Pneumatic Tubes Before beginning the installation of the tubes, a design was drafted in order to achieve the desired angles. Since the lane being studied has an additional width for loading/unloading passengers, traffic can sometimes drift into that area of the lane. The Interstate Highway standard for the U.S. Interstate

8 Highway System uses a 12 ft. (3.7 m) lane width (4), however observations and measurements needed to be made in order to determine the length of the tubes to be used in this survey. After watching traffic for a few minutes to establish a relative maximum width of the lane being utilized by typical traffic, a measuring tape was used to determine that an 18 ft. width from anchor to anchor across the lane would account for drifting traffic and vehicles with a wider wheelbase. Having established a lane width, calculations were done to ascertain that from the first anchors (closest to the modules), the second anchors needed to be offset at perpendicular distances of 1.57 ft. and 3.17 ft. to achieve the desired angles of 5 and 10. The vertical dashed lines represent a 0 offset of the first anchors, for which to measure the distances to place the second anchors, while maintaining a standard 3 ft. separation between anchors in each pair (Figure 6). 8 Figure 6: Design Plan for Tube Installation In order to install the tubes at these angles accurately, measurements and marks needed to be made on the road before laying the tubes downs and setting the anchors. To achieve true perpendicularity to the lane, a nearby crosswalk served as a reference point. First, the decided 18 ft. lane width was measured and marked on the crosswalk. Then, from the two marks on the cross walk, another decided 18 ft. was measured to where the first anchor points would be (Figure 7). These two points served as the basis of the subsequent measurements to achieve the desired angles. The measurements and marks shown in Figure 8: Measurements Before Installation were used as the anchor points for this survey. To minimize count differences as vehicles change in and out of the surveyed lane, the first anchors in each set of tubes were installed closely together with 1 ft. distance between pairs.

9 9 Figure 7: Crosswalk as 0 Reference Figure 8: Measurements Before Installation Once the anchors were set, each tube was stretched to a specific tension. According to the operator s guide provided by the manufacturer, 10% of the tube length is to be pulled through the webbing (3). This achieves a tension that prevents tires from dragging the tubes as cars pass by, but also does not

10 cause additional strain on the tubes. Since the lane being surveyed was 18 ft. wide, 1.8 ft. was pulled through the webbing wire. After plugging each tube into the appropriate connector on the modules and locking the modules to the nearby traffic signal pole, the tube installation was complete Figure NuMectrics NC-200 Figure 9: Completed Installation A NuMetrics NC-200 was installed in between the 0 tubes (Figure 10). This device detects vehicles by sensing the deviation in the earth s magnetic field as vehicles pass by. The NuMetrics is installed with a rubber cover and is screwed into the center of the lane. It is capable of recording traffic volumes, vehicle lengths, speeds, and more.

11 11 Figure 10: NuMetrics NC-200 Installed Baseline Data (Video) The primary basis for control volume was the CountCam system by Counting Cars, which records video from wide angled lenses for processing. The video is watched at variable speeds and each vehicle is counted by hand and classified as car/pickup truck or heavy vehicle using the CountPRO software (Figure 11) in combination with the DX-1 COUNTpad by Ergodex (Figure 12). Figure 11: CountPRO Software

12 12 Figure 12: Ergodex DX-1 COUNTpad The camera was fixed on top of a telescoping pole which was then attached to the traffic signal pole next to the tubes and pointed downward at all three sets of tubes with oncoming traffic in view (Figure 13).

13 13 Figure 13: CountCam System Installed Traffic was recorded continuously for the full 48-hour duration of the survey. 5. Data Analysis At the end of the survey, the data was extracted from each MetroCount module, the NuMetrics NC- 97, and the CountCam system. Using the MetroCount Traffic Executive software, several types of reports can be generated. These include Speed Reports, Vehicle Classification Reports, Vehicle Count Reports by either 15-minute or 1-hour interval, etc For the purposes of this study, the Individual Report was found to be most useful. The Individual Report provides a timestamp, speed, # of axles, and classification of each vehicle recorded throughout the survey, which can be exported to excel for further processing (Figure 14).

14 14 Figure 14: MetroCount Traffic Executive Software Furthermore, within this report, upon selecting a vehicle in the list, an Inspector Gadget tool appears which allows the user to examine the timing between triggers in milliseconds for each axle (Figure 15). Figure 15: Inspector Gadget Tool Displaying a Vehicle with 5 Axles While counting the video, select vehicles in the video were referenced to the Individual Report to verify timestamps and classifications. The same was done with reports generated by the NuMetrics NC-200. Counts by 15-minute interval were also referenced with the 0 tubes Vehicle Count Reports at the end of each interval in the 12 hours of video that was counted. This histogram compares the

15 counts by 15-minute interval between all three sets of tubes and the NC-200 before the video count was added (Figure 16). 15 Figure 16: Vehicle Counts by 15-Minute Interval Before Adding Video Count After adding the video count to the histogram, one can see that all devices measured traffic volumes with relative accuracy during low traffic times; from midnight until about 8 AM (Figure 17). Figure 17: Vehicle Counts After Adding Video Count According to this graph, the skewed tubes counted higher numbers than the straight tubes during high traffic times, as well occasionally during low traffic times. This is due to the skewed tubes detecting each wheel per axle separately, in turn causing the tube modules to register a single vehicle as two or more. This becomes evident when comparing the same vehicle between the three Individual Reports in Figure 18.

16 16 Figure 18: 5 Axle Semi-Truck Individual Report Comparison Note that the skewed tubes recorded higher speeds, which is likely another effect of each wheel being counted as an axle. The o s in the last columns are meant to represent the axle spacing per the vehicles classification, although the Coerced sequence flags means that the vehicle(s) were forced into the default/most common class. The video confirms that the 0 tubes were able to count and classify this vehicle correctly, but the skewed tubes divided the 5 axles of the semi-truck into 3 or 4 passenger cars with the same speed (Figure 19). Vehicles are classified in accordance with the current FHWA Vehicle Classification scheme (Figure 20)(6). Figure 19: 5 Axle Semi-Truck Recorded at 10:14:14 AM

17 17 Figure 20: FHWA Vehicle Classification Scheme (6) The inspector Gadget tool shows us what is happening on a miniscule level. Figure 21 displays the triggers for 5 axles on tubes A and B for the 0 pair, also showing the axle spacing relative to time in milliseconds.

18 18 Figure 21: 5 Axle Semi-Truck Inspector Gadget for 0 Tubes While the Inspector Gadget reads rather intuitively for the correctly installed tubes, when comparing the same vehicle on the skewed sets of tubes with this tool, the Inspector Gadget shows triggers for each wheel until the module produces the Coerced sequence flag (Figure 22 and Figure 23). Figure 22: 5 Axle Semi-Truck Inspector Gadget for 5 Tubes Figure 23: 5 Axle Semi-Truck Inspector Gadget for 10 Tubes Furthermore, as one may expect, all sets of tubes are able to count and classify motorcycles correctly, since they only have a single wheel per axle (Figure 24 and Figure 25). Note that the recorded speeds are more similar.

19 19 Figure 24: Motorcycle Individual Report Comparison Figure 25: Motorcycle Recorded at 10:52:40 AM Additionally, instead of scanning the Individual Report to determine if tubes were installed askew, the tubes may classify vehicles into classes that to not fit the demographic. Figure 26 shows the counts by class between the three sets of tubes over the first 24 hours of the survey.

20 20 Figure 26: Counts by Class Between All Three Sets of Tubes According to the 0 tubes and other traffic data specific to this demographic, classes 7, 10, 11, 12, and 13 are not typical. The fact that the skewed tubes classified at least one vehicle into one of these classes is another sign of a skewed tube installation. Also note that the number of vehicles in class 3 decrease while the number of vehicles in other classes increase in relation to the angle, while the number of motorcycles (class 1) remains the same in all three sets of tubes. Using the baseline data from the video count, the total error (Equation 1) was calculated for each set of tubes between the hours of 10 AM and 12 PM, when the skewed tubes appeared most troublesome. The 0 tubes generated a low total error of 0.85%, which was a difference of 4 passenger vehicles. The 5 tubes generated a high total error of 18.47%, while the 10 tubes generated a total error of 11.46%. During high traffic times such as this is when traffic data is most valuable, so total error as severe as those generated by the skewed tubes do not provide relevant information. Based on this information, tubes installed at angles of 5 or greater do not provide accurate or useful information about traffic volumes or vehicle classification, and repeating the survey is recommended if errors of this degree appear in the collected data. In the efforts to determine a threshold angle, at which point the data from the tubes may become unusable, calculations were inconclusive. Further testing of angles less than 5 are needed to provide a range of angles that produce acceptable traffic volume measurement and vehicle classification data. 6. Summary In conclusion, tubes installed at 0 provide the most accurate data, as expected, but achieving true perpendicularity to the surveyed lane(s) may involve an extra step during the installation process. As described in the data collection section, a point of reference was used to achieve a 0 offset between anchors. Alternatively, if no reference point is available, one can choose a first anchor point, then measure the shortest distance to a dividing/side line on the other side of the lane with a measuring tape as shown in Figure 27: How to Achieve Perpendicularity.

21 21 Figure 27: How to Achieve Perpendicularity There are ways to uncover a few tale tell signs of a skewed tube installation without having a baseline of comparison. Depending on the manufacturer/model of the tube modules and analysis software, an Individual Report or equivalent can be used to determine if repeating a survey is necessary. If the Individual Report shows multiple vehicles being counted with the exact same timestamp and speed, as shown in Figure 18, this is a sign that the tubes counting individual wheels. Using Excel, there is a simple way to determine if any duplicate timestamps exist. Select the timestamp column for a single day, then from the Data tab, select Remove Duplicates. This will generate a dialog window which states the number of duplicates, if any, and the number of unique values remaining (Figure 28). If any duplicates exist, undo the Remove Duplicates tool, since this tool removes the timestamps in question and consolidates the selected range. Then, to highlight these vehicles with duplicate timestamps, select the same range for a single day, then from the Home tab, select Conditional Formatting. From the dropdown list, select Highlight Cells Rules, then Duplicate Values. This will highlight all timestamps within the selected range with the same values. If more than one day is selected, this tool will highlight timestamps for vehicles that were counted at the same time on different days. Figure 28: Remove Duplicates Tool in Excel

22 Depending on the number of duplicate timestamps found, it is possible that the tubes were installed slightly askew, or some vehicles may have approached the tubes at an angle. If the tubes are spanned over multiple lanes and only a few duplicate timestamps are found, it is possible that two or more vehicles triggered the tubes at the same time, but the recorded speeds will most likely be different. In either case, investigation of the classification data is recommended to determine if the classifications appear feasible according to the region using one s best judgement. Since the skewed tubes registered certain vehicles into classes that do not fit the demographic, if a tube survey produces data with such occurrences as these, it is likely that the tubes were skewed and repeating the survey is recommended. It is likely that a small range of angles less than 5 exists that allow the tubes to count and classify vehicles with acceptable accuracy, but determining a threshold angle would involve more testing. 22

23 7. References 1) McGowen, Patrick, and Sanderson, Michael. "Accuracy of Pneumatic Road Tube Counters." (2011): Western District Annual Meeting Institute of Transportation Engineers. Web. 20 June ) Pell, Bob. "MetroCount Classifier Testing." AU Department of Infrastructure, Energy and Resources, 21 Jan Web. 20 July ) "MC5600 Roadside Unit Operator's Guide." MetroCount.com. MetroCount, Jan Web. 18 July ) "Mitigation Strategies For Design Exceptions - Safety Federal Highway Administration." U.S. Department of Transportation, 15 Oct Web. 19 July ) Peeta, Srinivas, and Pengchang Zhang. "Counting Device Selection and Reliability." Purdue University, Mar Web. 22 July ) "Vehicle Classification." Delaware Department of Transportation, n.d. Web. 21 July