Addressing Issues with GPS Data Accuracy and Position Update Rate for Field Traffic Studies

Transcription

1 Addressing Issues with GPS Data Accuracy and Position Update Rate for Field Traffic Studies THIS FEATURE VALIDATES INTRODUCTION Global positioning system (GPS) technologies have provided promising tools THE ACCURACY AND for transportation engineers to collect POSITION UPDATE RATE data such as vehicle trajectory, travel time and travel speed. 1,2 During data collection with a certain GPS device, the GPS PROVIDED BY THREE receiver communicates with satellites and DIFFERENT GPS RECEIVERS ground stations through its antenna. It outputs latitude, longitude and the Universal Coordinated Time, which can be FOR TRAFFIC STUDIES. GPS converted into travel time, distance, speed RECEIVERS WITH HIGH and delay. GPS devices also can improve the efficiency of traffic data collection and ACCURACY AND FAST RATE the safety of operators. 3 GPS receivers vary tremendously in USUALLY ARE EXPENSIVE features and price. To acquire quality traffic data, it is important to use select AND REQUIRE COMMERCIAL receivers with satisfied position accuracy and update rate. GPS satellite signals are DATA SERVICES. subject to some form of interference, such as selective availability (SA), atmospheric conditions and multipath. In the INTERPOLATION past, a regular positioning service for TECHNIQUES ARE PROVIDED civil use usually provided data with horizontal accuracy within 100 meters (m) FOR TRANSPORTATION with SA. After the use of SA was discontinued on May 1, 2000, regular GPS ENGINEERS TO ELIMINATE receivers could achieve accuracy within 10 m. Some receivers use differential THE DISCONTINUITY IN GPS GPS (DGPS), which improves accuracy by incorporating error corrections provided by a GPS monitoring station. DATA SO AS TO ENHANCE These receivers are expensive and require DATA QUALITY FOR TRAFFIC commercial data services. The Wide Area Augmentation System STUDIES. (WAAS) under construction should improve GPS accuracy to about 7 m. 4 WAAS consists of BY SHUO LI, P.E. AND KAREN ZHU, PH.D. approximately 25 ground reference stations, located in the United States, for correcting errors in GPS signal, timing and satellite orbit. It also has shown increased coverage and availability. Position update rate is another issue associated with the selection of GPS receivers. To avoid losing data around a position such as an intersection, transportation engineers may need continuous position data. Therefore, receivers should output position data continuously and rapidly. This feature discusses the selection of GPS receivers for field traffic studies. The analysis focuses on three typical receivers with different features and prices. Their accuracies were validated at a survey control station. Real travel time and speed data were measured using the receivers. Interpolation techniques were presented to construct missing data and improve data continuity. It is believed that if the goal of this study is fulfilled, transportation engineers will become more satisfied with the use of GPS devices, especially regular GPS receivers in field traffic studies. SELECTING A GPS RECEIVER To choose the right GPS receiver, transportation engineers should consider accuracy, position update rate, computer interface and price. GPS receivers vary tremendously in features and price, and it would be very difficult to validate all of the different receivers in this feature. For convenience, the GPS receivers for general civil use can be roughly divided into three categories. Category I includes receivers with submeter accuracy and fast position update rate, such as Trimble AgGPS These receivers usually use commercial DGPS data services and are expensive. Category II includes receivers with WAAS differential correction, such as Garmin GPS V. 6 They provide accuracy within 3 m and a position update rate of 1 Hertz (Hz). Category III includes regular receivers without differential correction, such as Pharos igps The accuracy for this category is about 10 to 15 m and the data rate is 1 Hz (or lower). No GPS receivers are designed specifically for field traffic studies. AgGPS 132 is designed for agriculture and GPS V is mainly for outdoor recreation. Different 32 ITE JOURNAL / FEBRUARY 2003

2 (a) Q 94 coordinates measured randomly (a) Position update rate with GPS receivers at rest Update rate (sec. per time) Time of testing (hh:mm:ss) (b) Q 94 coordinates measured in a 10-minute period (b) Position update rate with GPS receivers in traveling vehicles Update rate (sec. per time) Time of testing (hh:mm:ss) (c) Deviations in measured coordinates at Q 94 (c) Data rate after real-time processing Update rate (sec. per time) Figure 1. GPS coordinate measurements and deviations at station Q 94. studies may have different purposes and requirements. Therefore, it should be kept in mind that these accuracies and position update rates are assumed under ideal working conditions (e.g. a sufficient number of visible satellites, a wide and open environment, etc.). Furthermore, during data collection, the GPS receiver usually is installed in a moving car. Vehicle vibrations and rough pavement surfaces may cause the accuracy and position update rate to vary from time to time. VALIDATING POSITION ACCURACY AND POSITION UPDATE RATE Horizontal Position Accuracy A GPS receiver at rest provides better position accuracy than it does in a moving car. In this feature, GPS accuracy at rest was validated at survey control station Q 94, with North American Datum 83 coordinates of ' " (North) and ' " (West), in West Lafayette, IN, USA. 8 On May 15, 2002, its coordinates were measured using the three Figure 2. Variations of real-time position update rate with igps-180. selected GPS receivers in two scenarios. In the first scenario, each receiver was used to measure the coordinates randomly. In the second scenario, the three receivers were used to measure the coordinates continuously over a period of time. Figure 1 shows the measured data in terms of the Universal Transverse Mercator (UTM) coordinates in Indiana (UTM zone 16). During data collection, OmniStar differential correction was enabled for AgGPS 132, WAAS was enabled for GPS V and no differential correction was enabled for igps As shown in Figure 1(a), AgGPS 132 provided position data with good consistency and average accuracy of m. The average accuracy was m for GPS V and m for igps-180. In Figure 1(b), position data were collected continuously in two 10-minute periods (for example, from 11:30 a.m. to 11:40 a.m. and from 12:40 p.m. to 12:50 p.m.). Again, AgGPS 132 provided position data with the highest accuracy. Apparent inconsistency occurred among data collected using igps-180 during the two periods. ITE JOURNAL / FEBRUARY

3 (a) Travel time distance diagram (a) Travel time distance diagram Travel speed (km/h) Travel time (sec.) (b) Travel speed distance diagram Travel time (sec.) Travel speed (km/h) (b) Travel speed distance diagram (c) Trajectories while car entering the on-ramp (c) Average travel speed between intersections Average speed (km/h) Figure 3. Travel distance, speed, time and trajectories measured on Interstate 65. Figure 4. Travel time, distance and speed measured on U.S. Route 52. To further validate the effectiveness of WAAS differential correction in Indiana, Figure 1(c) shows the deviations in position data measured over a 10-minute period. During data collection, WAAS differential correction was enabled for both AgGPS 132 and GPS V. It is demonstrated that both receivers provided better consistency and accuracy than igps-180, which did not have differential correction. Position Update Rate Field traffic studies usually require continuous position data to avoid losing information at a location of interest. Many receivers with DGPS are designed to output position data at 1 Hz fast enough to provide good continuity. However, it is very difficult for regular GPS receivers to maintain the 1-Hz rate and, at times, they cannot output this data. This is mainly due to the coverage of GPS signals in a specific area and the technologies used in the receiver design. As shown in Figures 2(a) and 2(b), in general, no problems were observed with the position update rates for AgGPS 132 and igps-180, which were plotted using original GPS position data. Both receivers updated their position at 1 Hz. However, caution should be used when performing real-time data processing with igps-180 because the receiver outputs several default messages from the National Electrical Manufacturers Association (NEMA), such as GGA, GSA, GSV and RMC. Users have no option but to receive all of these messages. Because of the large amount of information, some data may be lost during real-time processing. For the AgGPS 132 receiver, users can change the selection of NEMA messages to output only one message such as RMC. Laptops with a speed of 300 MegaHertz (MHz) or higher are fast enough to process the information received. Figure 2(c) shows data rates generated by real-time data processing. It is demonstrated that no data were lost for AgGPS 132 but some data were lost for igps-180. The resulting data rate varied from 1 34 ITE JOURNAL / FEBRUARY 2003

4 to 5 seconds (sec.). To solve this problem, users should save the original igps-180 data and then execute post-processing. ENHANCING DATA CONTINUITY USING INTERPOLATION When a GPS receiver roves through forests, urban areas, or tunnels, GPS signals will be blocked by trees, buildings, or the tunnels themselves. This can cause lapses in the data received. In the authors opinion, interpolation techniques can be employed to remove those lapses and improve the continuity of data measured using regular GPS receivers. The discussion in this section is based on real data measured using igps-180 on Interstate 65 (I-65) and a coordinated signal system on U.S. Route 52 (US52), respectively. Measurements of Travel Time, Distance and Speed on I-65 The test section on I-65 (westbound), which is about three miles long, was located between State Road 26 (SR26) and SR25 in Lafayette, IN. To provide a reference for the data collected using igps-180, an AgGPS 132 receiver with OmniStar differential correction also was installed in the test car to collect the same data. Figure 3 shows the travel distance, speed, time and trajectories measured on I-65. As illustrated in Figures 3(a) and 3(b), the test car merged onto I-65 from SR26 and exited I-65 to SR25. Before entering the on-ramp, the test car was stopped by the first signal. Before exiting the off-ramp, the test car was stopped again by the second signal. The data measured on I-65 using igps-180 were very close to those measured using AgGPS 132. The distance between the first and third signals was 5,825.2 m as measured by igps-180 and 5,708 m as measured by AgGPS 132. The average travel speed was 55.7 kilometers/hour (km/h) for igps-180 and 54.6 km/h for AgGPS 132. Discrepancies occurred mainly on the on- and off-ramps when the test car made a sharp left or right turn. The reason is complicated; however, an effort can be made to lessen the possible discrepancy by securing the GPS receiver in the car. Figure 3(c) shows the trajectories of the test car while entering the on-ramp. Obviously, when igps-180 was secured in the car, it produced a trajectory very close to that measured using AgGPS 132. Measurements of Travel Time and Delay on US52 The second test route was a coordinated signal system on US52 in West Lafayette, IN. The signal system consisted of three intersections. Figure 4 shows the measured travel time, distance and speed. Again, the curves created using igps-180 data were close to those created with AgGPS 132 data. It is observed that during data collection, the test car was stopped at the first signal and the second signal, respectively. Table 1 gives the detailed travel time, stopped delay and average speed estimated from the data collected using the two GPS receivers. The travel time and stopped delay data produced by igps-180 were very close to those produced by AgGPS 132. Discrepancies can be observed between the average speeds. As discussed earlier, some information may be lost during real-time processing of igps-180 data; therefore, discontinuity occurred in the travel time and distance. Interpolation techniques also can be used to provide more Table 1. Travel time, delay and average speed measured using igps-180 and AgGPS 132 receivers. Travel Stopped Average time delay speed Segment Receiver (sec.) (sec.) (km/h) Before igps st signal AgGPS st signal to igps nd signal AgGPS nd signal to igps rd signal AgGPS continuous position data in this case and to produce better results such as average speed, as shown in Figure 4(c). Further discussion follows. Construction of Missing Data Using Interpolation To enhance the continuity of traffic data measured near trees, buildings and tunnels using regular GPS receivers, it is necessary to construct or predict the possible values for missing data. Transportation engineers can employ two interpolation techniques to construct missing data. The first technique is an interpolation algorithm based on the fundamental kinematics equations below a = V 2 V 1 t 2 t 1 (1) V = V 1 + at (2) S = S 1 + V 1 t at2 (3) where a denotes an acceleration, V denotes speed and S denotes distance. Subscripts 1 and 2 denote the beginning and ending points used for interpolation. For the kinematics interpolation, transportation engineers must compute acceleration with respect to the time and speed information at points with measured data. To illustrate how to use the kinematics interpolation, consider those data measured in Table 2, which were collected in the segment between the second and third signals (see Figure 4). The bold numbers are real data measured using igps-180. The other numbers are missing data constructed using interpolation. For example, during data collection, igps-180 did not output positions at 13 to 19 sec. Therefore, the procedure to construct acceleration, speed and distance is based on the speed and distance data at travel time (t) = 12 sec. and t = 20 sec. as follows Compute the time interval: t = 20 sec. 12 sec. = 8 sec. Compute acceleration using Eq. 1: a = km/h 36.8 km/h = 0.44 m/sec. 2 8 sec. ITE JOURNAL / FEBRUARY

5 Compute speed and distance at different times using Eqs. 2 and 3, respectively: y 1 = 26.4 km/h, y 2 = km/h, x 1 = 0 sec. and x 2 = 2 sec. At t = 1 sec., x = t = 1 sec. The speed at t = 1 sec. is At t = 19 sec., for example, t = 19 sec. 12 sec. = 7 sec. V = 36.8 km/h (0.44 m/sec. 2 )(7 sec.) = km/h S = m + (36.8 km/h)(7 sec.)/ (0.44 m/sec. 2 )(7 sec.) 2 = m The second interpolation technique is linear interpolation. This does not require computation of the acceleration at time t. Using linear interpolation, the missing speed and distance data can be computed directly from the equation below y = x x 2 y 1 + x x 1 y 2 x 1 x 2 x 2 x 1 All missing data constructed using linear interpolation also are given in Table 2. For example, consider the speed data during the travel time of 0 to 2 sec. Based on igps-180 data, Table 2. Data constructed using interpolation. Kinematics interpolation (4) Linear interpolation Travel Speed Distance Acceleration Average Distance Speed Average time (sec.) (km/h) (m) (m/sec. 2 ) speed (km/h) (m) (km/h) speed (km/h) y = 1 sec. 2 sec km/h + 1 sec. 0 sec km/h = km/h 0 sec. 2 sec. 2 sec. 0 sec. The results by the kinematics interpolation are similar to those by the linear interpolation because one assumes that the test car accelerates or decelerates uniformly during a specific period of time. If the interpolation technique is applied to the data measured on the whole test route, all missing data can be constructed and the discontinuity can be eliminated. When GPS receivers are used near tall buildings, it may create a problem known as the urban canyon. Tall buildings not only block GPS signals, but they also generate false data. This phenomenon is known as multipath. AgGPS 132 can be upgraded with Everest technology to enhance accuracy near buildings. 5 Most regular GPS receivers do not have this feature. However, interpolation techniques, especially the kinematics interpolation, can be utilized to correct those false data. Table 3 shows the original AgGPS 132 data collected when the test car was traveling on Ohio Street and approaching the intersection with Illinois Street in downtown Indianapolis, IN. Because of the buildings, some false data were received during the time period from 11:26:19 through 11:26:29. The X and Y coordinates are in UTM zone 16. It is clear that the test car was stopped before the intersection by the traffic signal. Based on the speed and time information at 11:26:19, 11:26:25 and 11:26:29, the vehicle movement was divided into two phases. Assuming that the test car decelerated uniformly in these two phases, respectively, the accelerations are computed below 11:26:19 to 11:26:25 11:26:25 to 11:26:29 a = 0.27 m/sec m/sec. = m/sec. 2 6 sec. a = 0.0 m/sec m/sec. = m/sec. 2 4 sec. The speeds and distances for those positions with false information can be computed using Eqs. 2 and 3 and are given in Table 3. To check the accuracy, the distance between the two points at 11:26:19 ( , ) and 11:26:29 ( , ) is m. The distance between these two points using the kinematics interpolation is m. CONCLUSIONS To choose GPS receivers for field traffic studies, transportation engineers need to consider features such as accuracy, position update rate and computer interface, and they must make a compromise between 36 ITE JOURNAL / FEBRUARY 2003

6 Table 3. Correction of false GPS data using kinematics interpolation. Original data Data after correction Time X (UTM 16) Y (UTM 16) Speed Distance Time Acceleration Speed Distance (hh:mm:ss) (m) (m) (m/sec.) (m) (hh:mm:ss) (m/sec. 2 ) (m/sec.) (m) 11:26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: :26: required features and price. GPS receivers with differential correction can meet the challenges of accuracy and position update rate in most field traffic studies. For example, AgGPS 132 with OmniStar differential correction provided a position accuracy within 1 m at 1 Hz. WAAS improved GPS V position accuracy to 2 to 4 m. Regular GPS receivers without differential correction, such as igps-180, are inexpensive and may achieve accuracy of 10 to 15 m or better. Because of availability and coverage of GPS signals, these receivers may not output position data at a consistent, specified rate and discontinuity may occur. On interstate highways, regular GPS receivers can measure travel time, distance and speed comparably to GPS receivers with differential correction. In a coordinated signal system near large objects such as trees or tall buildings, the false data or discontinuity in position data may generate significant deviations in travel distance and speed. However, interpolation techniques can be employed to eliminate the discontinuity and lessen the possible deviations. References 1. Li, S., K. Zhu, B.H.W. van Gelder, J. Nagle and C. Tuttle. Reconsideration of Sample Size Requirements for Field Traffic Data Collection Using GPS Devices. Accepted for publication in Transportation Research Record, Jiang, Y., and S. Li. Measuring and Analyzing Vehicle Position and Speed Data at Work Zones Using Global Positioning Systems. ITE Journal, Vol. 72, No. 3 (March 2002): Li, S., K. Zhu, B.H.W. van Gelder, J. Nagle and C. Tuttle. Improving Efficiency of INDOT Traffic Data Collection Using GPS Devices. SPR Final Report, Division of Research, Indiana Department of Transportation, WAAS Product Team, AND-730 Fact Sheets. Wide Area Augmentation System. gps.faa.gov/library/waas-f.htm (accessed June 3, 2002). 5. Trimble Navigation Limited. AgGPS 124/132 Operation Manual. Sunnyvale, CA, USA, Garmin International Corp. GPS V Personal Navigator Owner s Manual and Reference Guide. Kansas, USA, Pharos Science & Applications Inc. igps-180 Manual, Torrance, CA, USA, The National Geodetic Survey. The NGS Data Sheet. (accessed April 22, 2002). 9. OmniSTAR USA Inc. How It Works. how_techdesc.html (accessed April 30, 2002). SHUO LI, P.E., is a research engineer with the Division of Research in the Indiana Department of Transportation. His research interests include fundamental data for highway infrastructure studies, pavement NDT testing, pavement mechanistic analysis and smart pavement technologies. KAREN ZHU, Ph.D., is a senior system analyst with the Division of Research in the Indiana Department of Transportation. She received her Ph.D. from the University of California San Diego in Her research interests cover GPS-GIS data acquisition and management, advanced database and data acquisition for pavement instrumentation. ITE JOURNAL / FEBRUARY