Christopher M. Monsere TAC Meeting June 29, 2009 1:00-3:30PM

1. Using Existing ITS Commercial Vehicle Operation (ITS/CVO) Data to Develop Statewide (and Bi-state) Truck Travel Time Estimates and Other Freight Measures Christopher M. Monsere TAC Meeting June 29, 2009 1:00-3:30PM

2. Agenda

3. Objectives • Retrospectively study truck transponder data in key corridors to determine the feasibility of producing freight corridor performance measures. • Study the feasibility of using transponder data from commercial vehicles to predict corridor travel times with existing infrastructure • Demonstrate other freight performance measures

4. Data Almanac • 22 (20) reporting WIM sites in Oregon • All upstream of weigh stations • All are CVISN sites • PSU WIM Data Archive • April 2005 – February 2009 • 41,534,800 + trucks • Data quality • Intermittent data outages and problems • Accuracy ? • Focus of other research project

5. Axle weights • Gross vehicle weight • Axle spacing • Vehicle class • Bumper-to-bumper length • Speed • Unique transponder numbers WIM Layout Trucks Bypass R.F. Antenna 1 mile Notification Station (in-cab) High Speed Tracking System WIM Sorting Site Ramp (with Overhead AVI) Static Sorter Weighing System

7. Estimating Truck Travel Times • Matching transponders • Filtering through trucks • Results • Discussion

8. Trucks with Transponders, 2007

9. Defining Links • At each station, find all possible downstream stations. • Calculate shortest path between stations • Three categories • Primary • Route certain, one highway • Secondary • Route certain, more than 1 highway • Tertiary • Route uncertain

10. Primary – 211 ASP to BOR

11. Secondary – 234 KFP to LWL

12. Tertiary – 213 ASP to BND

13. Tertiary – 213 ASP to BND

14. 201 – FWB to EMH Free flow travel time = Distance / 55 mph = 126.4 mi / 55 mph = 2.3 hrs Search window = 2.3 *.75 = 1.7 hrs (74 mph) 2.3 * 2 = 4.6 hrs (27 mph)

15. Matched Trucks by Link

16. All Matched Trucks in Time Window FF

17. Filter Algorithm • For each truck j traveling on link i determine the estimated travel time, tj,i. • If the travel time tj,i is less than the free-flow time ffj,idenote this truck as a through truck. • If the travel time tj,i is less than the upper travel time utj,i (defined as an average travel time of 50 mph). • Find the median travel time mtj,i in the sample of X previous truck observations and compare that to tj,i. If tj,i does not exceed mtj,i by a threshold of Y, truck j is assumed to be a through vehicle. • If none of the above criteria are met, the tj,iis excluded (i.e., j is not a through truck).

18. Filtered Trucks (Green)

19. Through Trucks Only

20. 201 – FWB to EMH Ladd Summit RWIS El 3619 ft Pendleton NOAA El 1493 ft

21. Through Trucks and Temperature

22. Through Trucks and Rel. Humidity

23. Through Trucks and Weather

24. 201 – FWB to EMH, August 07

25. 201 – FWB to EMH, August 07

26. 201, Average Link Speed, by Day

27. Average Link Speed, by month

28. Results min 1000 obs

29. Results min 500 obs

30. 234 – KFP to LWL

31. 210 – LGR to ODF

32. Ground Truth Comparison • Questions • Can trucks estimate car travel times? • Are the truck matched times “reasonable”? • Method • State motor pool fleet • Low power GPS data logger Sample Data Tag,$GPGGA,UTC(hhmmss.sss),Latitude,N/S,Longitude,E/W,Fix quality,Number Of Satellites,Horizontal dilution of position,Altitude,Height of geoid,,Checksum Tag,$GPRMC,UTC(hhmmss.sss),A,Latitude,N/S,Longitude,E/W,Speed(knots),Course(degrees),Date(ddmmyy),,Checksum ---,$GPGGA,162807.000,3205.5748,S,11548.6228,E,1,46,226.6,7990.0,M,00.0,M,,*73 ---,$GPRMC,162807.000,A,3205.5748,S,11548.6228,E,0.00,46.00,080800,,*2B ---,$GPGGA,162807.000,3205.5749,S,11548.6228,E,1,46,226.6,8502.0,M,00.0,M,,*7A

33. Sample Mapped GPS data HH:MM:SS HH:MM:SS

34. Summary of Probe Data

35. Probe Trajectory, WDN to CSL

36. Probe Trajectory, ASP to BOR

37. Results

38. Discussion • Real time traveler information • Long term corridor performance • Data quality

39. Real time traveler information • Incidents and weather typical delays • Probe data confirms trucks behave differently than cars • Distance between stations • no vehicle conservation • increased in variability of travel times • latency in information • sensitivity of filter governs minimum incident that can be detected • Number of matches • See handout

40. Real time traveler information • Solve the spacing issue? • Additional tag readers ($9,000) • Where to place? • Junctions and other known delay areas • Minimum incident duration, link parameters • Still have trucks not like cars, especially winter • Still have data quality issues • Other approaches • MAC address matching • More passenger cars • Same issues

41. Long term corridor performance • Appears feasible • Sufficient number of matches • Average speeds and standard deviations easily calculated • Long distance not as problematic • Able to detect some disturbances • Additional readers at key locations could help with patterns • Work on data quality

42. Data quality • Generally, good • Outages • Planned/unplanned • Sensor error or drift • Weight • Timestamp • Could be addressed

43. Other Freight Performance Measures • Station level • Metrics produced from this category do not leverage the transponder information. These are traditional metrics that can be extracted from a WIM data source • Matched trucks • These are truck-pairs on each link observed within a window (not necessarily through trucks) • Filtered matched truck data • Travel time was the key measure (already presented)

44. Station Level • Counts of Trucks • Gross Vehicle Weight and Payload Estimate • Percentage of Trucks with Transponders • Overweight Vehicles • Observed Truck Speed

45. Counts of Trucks

46. Counts of Trucks

47. Overweight> 80 kips

48. Overweight> 105 kips

49. Matched Truck Data • Estimated Freight Activity on Corridor • Freight Patterns • Ton Miles • Emissions

50. Freight Activity Production About 1600 more truckloads consumed Consumption