View
3
Download
0
Category
Preview:
Citation preview
MODELLING LONG HAUL TRUCK ROUTE CHOICE IN ONTARIO
Smart Freight Centre Symposium27th November, 2020
Presenter: Syed Ubaid AliSupervisor: Dr. Kevin Gingerich
1
Outline
1. Introduction ‐> Goals & Objectives2. Modelling Method3. Data processing ‐> ArcGIS software4. Modelling ‐> C‐Logit models5. Model Performance6. Scenario Testing ‐> Travel time changes on major roadways7. Conclusions
2
Introduction: Freight Transportation In Canada• Approximately 112,000 trucking companies in Canada47,500 companies in Ontario (Statistics Canada, 2018)
1400150016001700180019002000210022002300240025002600270028002900300031003200
Average AA
DTT
Historical AADTT Year
Provincial Highway Network Historical Truck Volumes along Provincial Highways
3
AADTT grew almost 94% over 28 years
Research Goal
• Utilize existing patterns of long haul truck movements to develop a route choice model that:
explains and predicts long‐haul truck vehicle movements in Ontario
• Objectives Conduct a literature review on route modelling techniques and
route performance measures Devise an algorithm to generate route choice sets based on the
commonality factor. Process variables and estimate an advanced discrete choice
model.
Less Time
MoreRest Areas
MoreFreeways
4
Modelling Method
• Usage of Large Map‐matched GPS Datasets
• Define unique routes between a given origin and destination based on a commonality factor
• Test several combinations of descriptive variables to measure utility of a given route
5
Two Routes with no overlap; CF = 0%
Two Routes with partial overlap; CF = 50%
Two Routes with complete overlap; CF > 95%
𝐶𝐹𝑙𝐿 𝐿
KG3
Slide 5
KG3 One idea per slide. For example, you need to use the slide here to explain what 'map-matched GPS data' is.Kevin Gingerich, 6/3/2020
Utility Model (C‐Logit)
• C‐Logit model uses the CF to account for route overlap
𝑃exp 𝛽 𝑿𝒊𝟏 𝛽 𝑿𝒊𝟐 𝛽 𝑿𝒊𝟑 𝛽 .𝐶𝐹
∑ exp 𝛽 𝑿𝒋𝟏 𝛽 𝑿𝒋𝟐 𝛽 𝑿𝒋𝟑 𝛽∈ .𝐶𝐹
Where:• 𝑃 is the probability of a given decision maker selecting alternative i
• 𝛽 are parameters estimated by the model• 𝑋 are input variables• CF are commonality factors
6
𝑿𝒊𝟐 = Time
𝑿𝒊𝟏Rest Areas
𝑿𝒊𝟑Freeways
Research Data
• GPS Data• 50,431 Truck Trips• 840 OD‐Pairs• 13,700 US related trips• Collected over 1 week of March 2016
• Ontario Road Network Element Dataset (MTO)
7
Route Definition
• OD‐Pairs represent regions, cities, towns, etc. defined within census boundaries
• Routes represent the paths used to complete trips between a given OD‐Pair
• Trips that have high degree of overlap (CF) are grouped in to routes
8
PeelRegion
Ottawa
ArcGIS Model for Choice Set Generation
9
• Routes with CFs above the thresholdof uniqueness (e.g. 85%) are iterativelydeleted
Route Allocation
• After defining the route choice set, each trip is measured for its overlap with each route
• Routes are assigned back to the original trips
Trip No.113 chose Route No.3 due to high degree of overlap
10
Trips Routes
Model Applied to Entire Dataset of 50,431 trips
11
Model Run Time Initial Results Final Modelling Dataset
25 Hours
50,431 Trips
840 OD Pairs
2483 Unique Routes
37,111 trips
577 OD-Pairs
2,220 routes
263 OD-Pairs had on route option between them (no choice for driver to make)
These accounted for 13,320 Trips
Variables and Expectations
Usage of Freeways
12
Variable Description ExpectationAVGSPD Average speed in km/hr for a route based on observed GPS trips +TTMIN Minimum travel time for a route based on observed GPS trips ‐TTIME Average travel time in hours for a route based on observed GPS trips ‐TTMAX Maximum travel time for a route based on observed GPS trips ‐SIMSPD Average simulated speed for a route based on HERE data +
SIMTTMIN Minimum simulated travel time for a route based on HERE data ‐SIMTTIME Average simulated travel time for a route based HERE data ‐SIMTTMAX Maximum simulated travel time for a route based on HERE data ‐TTINDX Travel Time Index for a route based on observed GPS trips ‐BTINDX Buffer Time Index for a route based on observed GPS trips ‐PTINDX Planning Time Index for a route based on observed GPS trips ‐FWP Observed proportion of freeways comprising a given route +
FWP401 Observed proportion of Highway 401 comprising a given route +FWP4XX Observed proportion of freeways other than Highway 401 comprising a route +FWP407 Observed proportion of Highway 407 comprising a given route ‐INTRSCT Number of intersections along a given route ‐DIESEL Number of diesel gas stations along a given route +WEIGH Number of weigh stations along a given route +
CF Commonality Factor representing level of overlap for a given route and all otheralternatives +
C‐Logit Model Estimation
Variable Coefficient T‐StatisticMinimum Travel Time ‐1.67*** ‐77.95Freeway Proportion 1.18*** 22.56Proportion of Hwy401 1.97*** 42.67Number of Diesel Stations 0.27*** 57.18Number of Intersections ‐0.01*** ‐17.13Route Commonality Factor 0.69*** 11.62LL(0) ‐52495LL(β) ‐25022ρ2 0.523No. of Observations 37,111
13
*** indicates the parameter is statistically significant with 99% confidence
Model Performance
• Utility of a Route is given by• 𝑉 1.815620 𝑇𝑇𝑀𝐼𝑁 .984631 𝐹𝑊𝑃 .881460 𝐹𝑊𝑃401
.17326 𝐷𝐼𝐸𝑆𝐸𝐿 0.003510 𝐼𝑁𝑇𝑅𝑆𝐶𝑇 .36821 𝐶𝐹
• Utility is substituted into C‐Logit model to obtain probability of selecting a route in a choice set
• Route probability is multiplied by the observation frequency in an OD‐Pair to predict the number of trips per route
14
Observed v.s. Predicted Frequency Plot
15
R² = 0.9584
0100200300400500600700800900
1000
0 200 400 600 800 1000 1200
Pred
icted Freq
uency
Observed Frequency
Scenario Test: Congestion on Highway 401• TTMIN is expected to increase proportionally with FWP401
𝑇𝑇𝑀𝐼𝑁 𝑇𝑇𝑀𝐼𝑁 1 𝛼 𝐹𝑊𝑃401Where:• 𝑇𝑇𝑀𝐼𝑁 is the adjusted minimum travel time along a route identified for Scenario 1• 𝑇𝑇𝑀𝐼𝑁 is the current minimum travel time along a route in the current model dataset• FWP401 is the proportion of Highway 401 comprising a given route.• 𝛼 is a variable utilized to increase the TTMIN based on the proportion of Highway 401 for a given route.
• The value of 𝛼 is set to 1 to double the travel time along Highway 401.
• Applied to entire dataset of 34,625 trips, 1502 routes, 470 OD‐Pairs
• Future Route choice is predicted using MNL model
16
Scenario #1 Example
17
Existing Route Frequency Route Frequency after 401 is Congested
Waterloo to Battle Creek, MI
Macroscopic Analysis of Scenario
18
• OD‐Pairs with the highest change include:
• York Region and Ohio91% Trucks Rerouted
• Peel Region and Levis, QC97% Truck Rerouted
• Peel Region and Sherbrook, QC93% Rerouted
Conclusions
• Most important factors of route choice are:• Minimum Observed Travel Time ‐ (Most elastic)• Freeway Proportion +• Proportion of Highway 401 +• Number of Diesel Stations Along route +• Number of Intersections along Route –
19
Research Implications
• The algorithms developed can be applied to any truck trip dataset
• The choice set generation process can immediately revealthe adequacy of connectivity between zones
• Route choice model factors can help develop adequateroutes in the future
• Collision data can be overlaid onto routes identified to provide initial screening of dangerous routes for long‐haul trips
• New or existing routes can be assessed to forecast future travel patternsfor long‐haul trucks by applying appropriate growth statistics
20
References• Ben‐Akiva, M., Toledo, T., Santos, J., Marzano, V., Lee, Y. J., Zhao, F., & Cox, N. (2015). Freight data collection using GPS and web‐based surveys: Insights from US truck drivers’ survey and perspectives for urban freight. Case Studies on Transport Policy, 4, 1–15.• Bovy, P. H. L., Bekhor, S., & Prato, C. G. (2008). The Factor of Revisited Path Size ‐ Alternative Derivation. Transportation Research Record, 2076, 132–140. Retrieved from https://journals.sagepub.com/doi/10.3141/2076‐15• Cascetta, E., Russo, F., Viola, F. A., & Vitetta, A. (2002). A model of route perception in urban road networks. Transportation Research Part B, 36, 577–592.• Cascetta, E., Russo, F., & Vitetta, A. (1997). STOCHASTIC USER EQUILIBRIUM ASSIGNMENT WITH EXPLICIT PATH ENUMERATION: COMPARISON OF MODELS AND ALGORITHMS. In 8th IFAC/IFIP/IFORS Symposium on Transportation Systems 1997 (TS ’97), Chania, Greece, 16‐18 June
(pp. 1031–1037).• Dalumpines, R., & Scott, D. M. (2011). GIS‐based Map‐matching: Development and Demonstration of a Postprocessing Mapmatching Algorithm for Transportation Research. Advancing Geoinformation Science for a Changing World, 1, 101–120.• Freitas, L. M. de. (2018). A recursive logit multimodal route choice model. Swiss Federal Institute of Technology Zurich. Retrieved from https://ethz.ch/content/dam/ethz/special‐interest/baug/ivt/ivt‐dam/publications/students/501‐600/sa600.pdf• Gingerich, K., & Maoh, H. (2017). Big Data Analysis to Measure Delays of Canadian Domestic and Cross‐Border Truck Trips. In Transportation Research Board 96th Annual Meeting. Retrieved from https://trid.trb.org/view/1437678• Gingerich, K., & Maoh, H. (2019). The role of airport proximity on warehouse location and associated truck trips: Evidence from Toronto, Ontario. Journal of Transport Geography, 74, 97–109. Retrieved from https://doi.org/10.1016/j.jtrangeo.2018.11.010• Jan, O., Horowitz, A. J., & Peng, Z.‐R. (2000). Using GPS Data to Understand Variations in Path Choice. Journal of the Transportation Research Board, 1725(1), 37–44.• Kaneko, N., Oka, H., Chikaraishi, M., Becker, H., & Fukuda, D. (2018). Route Choice Analysis in the Tokyo Metropolitan Area Using a Link‐based Recursive Logit Model Featuring Link Awareness. Transportation Research Procedia, 34, 251–258.
https://doi.org/10.1016/j.trpro.2018.11.039• Knorring, J. H., He, R., & Kornhauser, A. L. (2005). Analysis of Route Choice Decisions by Long‐Haul Truck Drivers. Journal of the Transportation Research Board, 1923, 46–60.• Kunchev, L. (2017). Methodology for selection the truck route. In Engineering for Rural Development (Vol. 16, pp. 263–272). https://doi.org/10.22616/ERDev2017.16.N052• Li, J., & Lai, X. (2019). Modelling travellers’ route choice behaviours with the concept of equivalent impedance. Transportation, 46, 233–262.• Luong, T. D., Tahlyan, D., & Pinjari, A. R. (2018). Comprehensive Exploratory Analysis of Truck Route Choice Diversity in Florida. Transportation Research Record. https://doi.org/10.1177/0361198118784175• MTO iCorridor. (2019). Historical AADT & AADTT. Retrieved May 11, 2020, from https://www.arcgis.com/home/item.html?id=https://services.arcgis.com/6iGx1Dq91oKtcE7x/arcgis/rest/services/Historical_AADT/FeatureServer• Oka, H., Hagino, Y., Kenmochi, T., Tani, R., Nishi, R., Endo, K., & Fukuda, D. (2017). Predicting travel pattern changes of freight trucks in the Tokyo Metropolitan area based on the latest large‐scale urban freight. Transportation Research, Part E, 1–20.• Ontario Data. (2015). Ontario Road Network: Road Net Element. Retrieved May 11, 2020, from https://data.ontario.ca/dataset/ontario‐road‐network‐road‐net‐element• Prato, C. G. (2009). Route choice modeling: past, present and future research directions. Journal of Choice Modelling, 2, 65–100.• Prato, C. G., & Bekhor, S. (2007). Modeling Route Choice Behavior: How Relevant Is the Composition of Choice Set? Transportation Research Record, 2003, 64–73. Retrieved from https://journals.sagepub.com/doi/10.3141/2003‐09• Ramming, M. S. (2002). Network Knowledge and Route Choice. Massachusetts Institute of Technology. Retrieved from https://dspace.mit.edu/handle/1721.1/49797• Rowell, M., Gagliano, A., & Goodchild, A. (2014). Identifying truck route choice priorities: the implications for travel models. Transportation Letters, 6, 98–106.• Scholars Portal. (2012). Scholars GeoPortal. Retrieved from http://geo2.scholarsportal.info/• Sobhani, A., Aliabadi, H. A., & Farooq, B. (2019). Metropolis‐Hasting based Expanded Path Size Logit model for cyclists’ route choice using GPS data. International Journal of Transportation Science and Technology, 8(2), 161–175.• Statistics Canada. (2018a). General Freight Trucking ‐ 4841. Retrieved June 20, 2019, from https://www.ic.gc.ca/app/scr/app/cis/summary‐sommaire/484• Statistics Canada. (2018b). Vehicle registrations. Retrieved from https://www150.statcan.gc.ca/n1/daily‐quotidien/180615/dq180615e‐eng.ht• Statistics Canada. (2020). Retail e‐commerce sales, unadjusted (x 1,000). Retrieved from https://doi.org/10.25318/2010007201‐eng• Tahlyan, D., Luong, T. D., Pinjari, A. R., & Ozkul, S. (2017). Development and Analysis of Truck Route Choice Data for the Tampa Bay Region using GPS Data.• Tahlyan, D., Pinjari, A. R., Luong, T. D., & Ozkul, S. (2017). Truck Route Choice Modeling using Large Streams of GPS Data. Tampa FL. Retrieved from https://rosap.ntl.bts.gov/view/dot/34888• Telgen, M. G. (2010). Realistic route choice modeling. University of Twente. Retrieved from https://essay.utwente.nl/59718/1/MA_thesis_M_Telgen.pdf• Transport Canada. (2017). Transportation in Canada 2016. Retrieved from https://www.tc.gc.ca/media/documents/policy/comprehensive_report_2016.pdf• Wang, Z., & Goodchild, A. V. (2014). GPS Data Analysis of the Impact of Tolling on Truck Speed and Routing. Journal of the Transportation Research Board, 2411, 112–119.
21
This research has been accepted for presentation at the upcoming Transportation Research Board Conference
Recommended