Development of a Tour-Based Truck Travel Demand Model

Transcription

1 Development of a Tour-Based Truck Travel Demand Model presented to 2014 Innovations in Travel Modeling presented by Cambridge Systematics, Inc. Arun Kuppam coauthored by Jason Lemp, CS Dan Beagan, CS Vladimir Livshits, MAG Lavanya Vallabhaneni, MAG Sreevatsa Nippani, MAG April 29, 2014

2 Overview Why? Who? What? How? Where? Truck trip diary surveys are expensive, with poor response rates Third-party commercial GPS vendors Deploy GPS units and collect GPS records from truck fleets Process GPS data to obtain truck trips and tours CS applied in Los Angeles, Chicago, Phoenix, Baltimore Example focuses on specifics from the Phoenix experience 2

3 Commercial Vehicle GPS Data GPS devices are widely deployed in cell phones, autos, and trucks Truck fleet operators subscribe to GPS services for operational and maintenance purposes Automatic Vehicle Location (AVL); Events Activated Tracking Systems (EATS); Fleet Telematics Systems (FTS) Entities (e.g., ATRI) collect and store historical GPS from operators (2004 to present) Condition of dissemination is that truck ID s are anonymized GPS information should be processed before it can be used for truck travel models 3

4 ATRI Data Specifics Heavy Trucks Large Sample of Trucks Relatively Cheap Cost for one month of data $10K ATRI was purchased to supplement heavy truck (FHWA Classes 8-13) trip diary surveys 3.5 million positional records from 22,000 trucks Close to 60,000 truck tours 4

5 Truck GPS Data from Phoenix All Trucks in April 2011 ATRI GPS All Truck IDs April 2011 All Trucks April 2011 GPS Events 3,429,603 Truck Trip Ends?? Truck Tours 58,637 Trucks 22,657 5

6 Truck GPS Data from Phoenix One Truck in April 2011 ATRI GPS Truck ID April 2011 One Truck (ID ) April 2011 GPS Events 719 Truck Trip Ends?? Truck Tours 40 Trucks 1 6

7 Truck GPS Data from Phoenix One Truck on April 1, 2011 ATRI GPS Truck ID April 1, 2011 One Truck (ID ) April 1, 2011 GPS Events 25 Truck Trip Ends?? Truck Tours 1 Trucks 1 7

8 Processing of One Truck Tour Primary Anonymized Data Processed Data Trip Distance (miles) Time From Time Speed From Speed To Truck_ID Date Time Longitude Latitude Event Last Next Last (min) To Next (min) Last (mph) Next (mph) Event Type TAZ LU /1/2011 3:47:08 A.M First Starting 884 Industrial /1/2011 4:00:00 A.M moving /1/2011 4:02:38 A.M moving /1/2011 4:13:44 A.M Stopping /1/2011 4:14:00 A.M stopped 413 Industrial /1/2011 4:16:06 A.M stopped 413 Industrial /1/2011 4:22:36 A.M stopped 413 Industrial /1/2011 4:23:14 A.M stopped 413 Industrial /1/2011 4:23:28 A.M stopped 413 Industrial /1/2011 5:02:38 A.M stopped 413 Industrial /1/2011 5:22:38 A.M stopped 413 Industrial /1/2011 5:30:00 A.M Starting 413 Industrial /1/2011 5:45:00 A.M moving /1/2011 6:00:00 A.M moving /1/2011 6:01:14 A.M Stopping /1/2011 6:01:24 A.M stopped 745 Landfill /1/2011 6:02:38 A.M stopped 745 Landfill /1/2011 7:23:14 A.M stopped 745 Landfill /1/2011 7:33:38 A.M stopped 745 Landfill /1/2011 7:34:30 A.M Starting 745 Landfill /1/2011 7:45:00 A.M Moving /1/2011 7:48:54 A.M Stopping 884 Industrial /1/2011 8:02:40 A.M Last Stopped 8

9 Processing of One Truck Tour (continued) Primary Anonymized Data Processed Data Trip Time Speed Speed Distance From Time From To Truck_ (miles) Last To Next Last Next Event ID Date Time Longitude Latitude Event Last Next (min) (min) (mph) (mph) Type TAZ LU 4/1/ 3:47:08 First Industrial A.M. Starting One Truck April 1, One Truck April 1, 4/1/ 3:47: (ID ) Stopping 2011 A.M (ID ) 2011 GPS Transition Events 6 GPS Events 3 4/1/ 3:47:08 Industrial Truck Starting Trip Ends A.M. 4 Truck Trip Ends 3 Truck Tours 1 Truck Tours 1 4/1/ 3:47: Stopping Trucks 2011 A.M. 1 Trucks 1 4/1/ 3:47: Starting A.M. Land-fill 4/1/ 3:47:08 Industrial Stopping A.M. 9

10 Truck GPS Data from Phoenix Processing of One Truck on April 1, 2011 Processed One Truck April 1, 2011 Truck Trip Ends 3 Truck Tours 1 Trucks 1 ATRI GPS Truck ID April 1, 2011 Actual Stops 10

11 Truck GPS Data from Phoenix TAZ of Trip Ends for One Truck on April 1, 2011 Processed One Truck April 1, 2011 Truck Trip Ends 3 Truck Tours 1 Trucks 1 11

12 Truck GPS Data from Phoenix LU of Trip Ends for One Truck on April 1, 2011 Industrial Processed One Truck April 1, 2011 Truck Trip Ends 3 Truck Tours 1 Trucks 1 Landfill, Sand & Gravel Industrial 12

13 Truck Tour-Based Model Structure Tour Generation Stop Generation Tour Completion Stop Purpose Stop Location Stop TOD Choice Heavy truck tour rates by industry type 1 stop 2 stops Yes return to home base No does not return One of 10 stop types Retail Construction Farming Residential Government Warehouse Transportation Office Industrial Service One of 3,000 TAZs 1 st Stop TOD (24 one-hour periods) Next Stop TOD (24 one-hour periods) 11 stops 13

14 Stop Generation Model Predicts Number of Stops on Each Truck Tour Available Set of Choices Decision-Making Variables Starting LU of the tour Employment and households at starting TAZ Accessibility to employment Number of Stops by Industry Type Model Structure = MNL Utility Constr constr Farm farm Govt govt Ware ware Ind ind Other other -7 Number of Stops Outputs = Number of stops made on tour (any value between 1 and 11) 15

15 Tour Completion Model Predicts if Truck Returns to Home Base Available Set of Choices Decision-Making Variables Number of stops by industry type Total employment in starting TAZ Employment by industry type Model Structure = Binary Logit Utility Tour Incompletion by Number of Stops Farm Hhld Govt Ware Ind Ret Other Number of Stops Outputs = Tour is complete or not 16

16 Stop Purpose Model Predicts Purpose of Stop Available Set of Choices Decision-making Decision-Making Variables Previous purpose type Accessibility to employment No of stops on tour Tour purpose type Model Structure = MNL Log (1 + stop sequence number) Outputs = 1 of 10 stop purpose types 16

17 Stop Location Model Predicts Location TAZ of Each Stop Available Set of Choices Decision-making Decision-Making Variables Travel time between stops No of stops by type and Employment by type Accessibility to employment Model Structure = MNL Utility Impact of Travel Time on Utility Travel Time (Minutes) 1 Stop Tour 2 Stop Tour 3 Stop Tour 5 Stop Tour 8 Stop Tour 11 Stop Tour Outputs = Location TAZ of each stop 18

18 Stop TOD Choice Model Predicts TOD of each stop Available Set of Choices Decision-Making Variables Indicator for tour completion Travel time between stops Number of stops on tour Departure shift variable Model Structure = MNL Time/number of stops remaining Outputs = 1 of 24 hour intervals 19

19 Initial Findings Number of Stops Number of Stops by Tour Type 10% TOD of First Stop 3.0 8% % 1.5 4% % 0.0 0% % TOD of Subsequent Stops Percent of Stop Purposes Stop Purpose Distribution 8% 35% 6% 30% 25% 4% 20% 15% 2% 10% 5% 0% % Ret Cns Frm Min Hhs Gov War Trn Off Ind Srv Stop Purposes Model Target 19

20 Next Steps Perform more rigorous model calibration and validation Determine adequacy of GPS data both in terms of sample sizes and biases Implement truck tour model and compare with trip-based model applications Acquire GPS data for medium trucks and develop a medium truck tour model 20

21 Conclusions Truck model parameters are difficult and expensive to obtain by surveys Truck tourbased models require large samples by industry type Truck GPS data is a cost-effective way to obtain large volume of information from trucks Tour-based models capture trip chaining patterns of trucks by type 21