MODELING COMMERCIAL VEHICLE DAILY TOUR CHAINING

Jane Lin Ph.D., Associate Professor,

Department of Civil and Material EngineeringUniversity of Illinois at Chicago

SHRP 2 Freight Data and Modeling SymposiumSeptember 14, 2010

System ID Simulation Theory Hardware Discussion 2

• The focus of this study is on urban/regional commercial vehicle movements as opposed to long-haul freight movements.

• Urban commercial vehicle movements in this study include both goods and service deliveries.

• Definitions:– Trip: travel between two consecutive stops,– Base Stop: a stop where a tour originates and terminates,

which is typically a distribution center, a warehouse or the company office/garage where the vehicle belongs,

– Tour: a tour is made when a vehicle makes one or multiple stops before returning to a base stop,

– Daily Tour Chain: the entire movements of the commercial vehicle in a given day.

Study Scope And Definitions

System ID Simulation Theory Hardware Discussion 3

• Tour-based approach to modeling the underlying logistics of commercial vehicle tour strategies.

• Tour-based strategies:Burns et al. (1985)– Direct: a vehicle ships separate loads directly to

each customer without intermediate stops,– Peddling: a vehicle delivers to more than one

customer per load, making multiple stops before returning to the base stop.• Good for smaller quantities of higher value items, larger

truck capacity and higher customer density.

State-of-the-Art Urban Commercial Vehicle Modeling

System ID Simulation Theory Hardware Discussion 4

• Tour-based strategies:Liu et al. (2003)– Direct: suppliers operate independently– Hub-and-spoke: consolidates from multiple

suppliers and collectively delivers to customers.• Other tour-based studies

Holguin-Veras and Patil (2005)– Trip chaining was characterized by # trips chained,

length, and trip purposeWang and Holguin-Veras (2008)– Tour destination choice modelingHolguin-Veras and Thorson (2003)– Empty trip modeling with a tour

State-of-the-Art Modeling (cont’d)

System ID Simulation Theory Hardware Discussion 5

• Individual tour-based approach does not capture the interrelationship among linked tours (i.e., tour chain) and thus fails to incorporate tour chaining decisions and daily logistic planning.

• Some studies found that almost one of four commercial vehicles made more than one tour per day (Holguin-Veras and Patil 2005)

• The planning process of a single tour or linked tours during a daily operation involves many factors like customer demand, location, commodity type and shipment requirements, and logistics costs.

• Among the existing urban commercial vehicle studies the authors have reviewed, very few looked at tour chaining patterns.

Some Reflections on Tour-based Approach

System ID Simulation Theory Hardware Discussion 6

Observed Commercial Vehicle Daily Tour Chaining Strategies

(1) Single Direct (2) Single Peddling(3) Multiple Direct with One/Multiple Base Location

……… ……… ………

………

(a) (b)

(b)

………

………

………

(a)

………

(4) Multiple Peddling with One/Multiple Base Location

………

(a) (b)

(5) Mixed with One/Multiple Base Location

……… ………

………Base Location

Stops

System ID Simulation Theory Hardware Discussion 7

• Using the above tour chaining definition – to understand the daily operations of urban

commercial vehicles for both goods and service deliveries

– to investigate what characteristics of the delivery activity are associated with and possibly lead to a particular tour chaining strategy

Research Questions and Objectives

System ID Simulation Theory Hardware Discussion 8

• Solicit urban commercial vehicle survey data– Texas Commercial Vehicle Surveys in San

Antonio, Amarillo, Valley, Lubbock and Austin during  2005 and 2006

• Data exploration• Daily tour chaining strategy choice model• An individual tour-based choice model is also

built using the same data and variables– to test our hypothesis that tour chaining approach

better captures the holistic decision on commercial vehicle daily operation.

Research Approach

System ID Simulation Theory Hardware Discussion 9

Texas Commercial Vehicle Surveys

• San Antonio, Amarillo, Valley, Lubbock and Austin in  2005 and 2006• A total of 13,802 trips made by 1,711 commercial vehicles.  

Stop-level attribute DescriptionLongitude and latitude Stop coordinatesDeparture and arrival time Departure and arrival time at stopCargo type 22 Cargo Classifications: 1) Farm products, 2) Forest products, 3) Marine

Products, 4) Metals and Minerals, 5) Food, Health, and Beauty Products, 6) Tobacco Products, 7) Textiles, 8) Wood Products, 9) Printed Matter, 10) Chemical Products, 11) Refined Petroleum or Coal Products, 12) Rubber, Plastic, and Styrofoam Products, 13) Clay, Concrete, Glass, or Stone, 14) Manufactured Goods/Equip, 15) Wastes, 16) Miscellaneous Shipments, 17) Hazardous Materials, 18) Transportation, 19) Unclassified Cargo, 20) Driver Refuse to Answer, 21) Unknown to Driver, and 22) Empty

Cargo weight Total loaded or unloaded cargo weightActivity type 1) Base Location/Return to Base Location, 2) Delivery, 3) Pick-up, 4) Pick-

up and Delivery, 5) Maintenance (fuel, oil, etc.), 6) Driver Needs (lunch, etc.), 7) To Home, 8) Others (specify), and 9) Refused/Unknown

Land use type 1) Office Building, 2) Retail/Shopping, 3) Industrial/Manufacturing, 4) Medical/Hospital, 5) Educational (12th Grade or less), 6) Educational (College, Trade, etc.), 7) Government Office/Building, 8) Residential, 9) Airport, 10) Intermodal Facility, 11) Warehouse, 12) Distribution Center, 13) Construction Site, 14) Others (specify), and 15) Refused/Unknown.

System ID Simulation Theory Hardware Discussion 10

Data Exploration

Count (%) Number of Tours

within Daily Tour Chain

Number of Stops within

Individual Tour

Number of Stops within

Daily Tour Chain Mean Mean Mean

Single Direct 293 (17.12%) 1.00 1.00 1.00 Single Peddling 858 (50.15%) 1.00 7.16 7.16 Multiple Di rect 255 (14.90%) 3.40 1.00 3.40 Multiple Peddling 120 (7.01%) 2.31 4.51 9.98 Mixed 185 (10.81%) 3.12 1.85 5.56

System ID Simulation Theory Hardware Discussion 11

Data Exploration (cont’d)

Single Direct

Single Peddling

Multiple Direct

Multiple Peddling Mixed

Count 293 858 255 120 185 Avg. Mileage (Std.Dev.)

1405.4 (2644.4)

595.4 (1110.0)

591.1 (1147.9)

392.9 (796.2)

549.6 (975.3)

Avg. Max. Mileage (Std.Dev.)

401.3 (1602.2)

903.9 (2385.8)

676.2 (2092.8)

749.9 (2195.6)

1057.6 (2566.7)

Avg. Dwell (Std.Dev.)

262.5 (201.6)

46.6 (51.9)

55.3 (38.8)

30.8 (19.4)

42.7 (34.6)

% serving distribution centers 16.04 30.42 7.06 32.50 21.62 % serving retail 22.2 50.2 10.6 38.3 40 % Pick-up 3.07 31.58 53.72 35.83 45.94 % shipping cargo 13 5.11 8.39 43.92 11.66 18.37 % shipping cargo 14 26.62 19.81 10.58 23.33 20 % shipping cargo 16 6.14 12.12 1.17 15.83 4.32 % containing external trips 17.06 20.97 10.19 5.83 18.37

System ID Simulation Theory Hardware Discussion 12

• After initial investigation, a multinomial logit model is found appropriate• Five alternative strategies: single direct, single peddling, multiple

direct, multiple peddling, and mixed• Possible explanatory variables

– Land use type– Cargo type– Trip purpose– Travel distance between stops– Average dwell time– Average cargo weight– Truck type– External trip– Socio-economic, demographic, and transportation statistics

• Employment, Population, Household density, Median income, VHT, VMT, PMT, PHT in TAZ; etc.

Daily Tour Chaining Strategy Choice Model

System ID Simulation Theory Hardware Discussion 13

Explanatory VariablesVariable Description

Travel Distance Min Mile Minimum travel distance between stops within a daily tour chain Max Mile Maximum travel distance between stops within a daily tour chain

Base Mile Average distance between the base and the first or last stop within a daily tour chain

Total Mile Total mileage of a tour chain Avg. Mile Average mileage of a tour chain Type of Place (binary variable) Office/Retail/Manufacture/Residential/Warehouse/Distribution/Construction

=1 if a tour chain includes at least one stop at the type of place

Type of Cargo (binary variable) Cargo 1 thru 22 =1 if the specific cargo is shipped Trip Purpose (binary variable) Drop-off/Pick-up/Service =1 if a stop is drop-off/pick-up/service Other Operational Variable Avg. Dwell Average dwell time (10-3 mins) at stops in a tour chain Avg. Weight Average cargo weight (10-3 lbs) in a tour chain External =1 if a tour chain includes an external trip Semi Truck =1 if the vehicle is a semi truck (all Tractor-Trailer Combinations) Area Transportation, Economics and Demographics Employment counts by 2-digit NAICS, VHT, VMT, PMT, PHT; Tota l population, household density, median income by TAZ CBP employment counts by 2-digit NAICS by county Employment counts by 2-digit NAICS by MPO MSA area size

System ID Simulation Theory Hardware Discussion 14

Daily Tour Chaining Strategy Choice Model Result

Alternative-Specific Variables Coefficient Standard DeviationCoefficient Single Peddling 1.3155*** 0.3685 Multiple Direct 1.0934** 0.5015 Multiple Peddling 1.2518** 0.5834 Mixed -0.0593 0.4572 Goods Pickup From A Stop Other Than The Base Location Single Peddling 1.8330*** 0.3738 Multiple Direct 2.6359*** 0.3951 Multiple Peddling 1.5613*** 0.4245 Mixed 2.3424*** 0.4004 Average Distance (10-3 mile) the Vehicle Travel Between Stops Visited Single Peddling -0.1975*** 0.0582 Multiple Direct -0.2315*** 0.0865 Multiple Peddling -0.5193*** 0.1553 Mixed -0.2826*** 0.0942

* Significant at 10%. ** Significant at 5%. *** Significant at 1%

Log-Likelihood at Constant/Convergence -1956.7071/ -1526.345

Rho-Squared w.r.t Constant/Adjusted 0.21994/ 0.21499

Number of Observations 1428

System ID Simulation Theory Hardware Discussion 15

Daily Tour Chaining Strategy Choice Model Result (cont’d)

* Significant at 10%. ** Significant at 5%. *** Significant at 1%

Alternative-Specific Variables Coefficient Standard DeviationAverage Dwell Time (10-3 mins) at Stop Single Peddling -13.0966*** 1.0936

Multiple Direct -9.2286*** 1.3863

Multiple Peddling -24.5213*** 4.1375

Mixed -14.6342*** 2.1567

Shipment of Clay, Concrete, Glass, Or Stone (Cargo 13) Multiple Direct 1.4690*** 0.3695

Shipment of Manufacturing Goods/Equip. (Cargo 14) Single Peddling -0.83403** 0.3476

Shipment of Miscellaneous (Cargo 16) Multiple Direct -1.6838** 0.7035

Mixed -1.0570** 0.5291

At Least One Stop At Distribution Centers Multiple Peddling 0.6248* 0.3377

At Least One Stop At External Locations Single Peddling 0.9027*** 0.2966

Mixed 0.7499** 0.3537

System ID Simulation Theory Hardware Discussion 16

• Single direct is less likely to be used for tours with pick-up assignments in non-base location or with closely located stops than other strategies.

• Vehicles operating in a single direct strategy have longer dwelling time at stops than those in other strategies.

• Single peddling is preferred for external trips and less attractive when shipping manufacturing products is involved.

• Tours that involve shipments of Clay, Concrete, Glass, or Stone are more likely to operate in the multiple direct strategies, whereas those with shipments of miscellaneous products are less likely so.

• Tours that include at least one stop at a distribution center tend to run in multiple peddling.

• The use of the mixed strategies will increase if external trips are involved, and decrease if miscellaneous products shipments are involved.

Summary Findings of Daily Tour Chaining Model

System ID Simulation Theory Hardware Discussion 17

• A binary logit model is built• Two alternative strategies: direct and peddling

• Same list of possible explanatory variables as seen previously only at the tour level– Land use type– Cargo type– Trip purpose– Travel distance between stops– Average dwell time– Average cargo weight– Truck type– External trip– Socio-economic, demographic, and transportation statistics.

Individual Tour-based Choice Model

System ID Simulation Theory Hardware Discussion 18

Tour-based Choice Model ResultsVariables Coefficient Standard Deviation

Constant -0.8384*** 0.2276Min. Mile -0.5165** 0.2121Distribution Center 0.9423*** 0.1811Retail Stores 1.2224*** 0.1561Miscellaneous Cargo (Cargo 16) 0.7410** 0.3018Transportation Cargo (Cargo 18) 1.1448*** 0.3130Pick-up 3.9849*** 0.2974Avg. Dwell (10-3 minutes) -0.5458*** 0.7907Avg. Weight (10-3 lbs) -0.0961*** 0.0063External 1.1376*** 0.2275Semi Truck 0.8116*** 0.1988Household Density in TAZ (10-3 per Sq. miles) 0.1518** 0.0682Log-Likelihood at Constant -1449.1878Log-Likelihood at Convergence -763.89687Rho-Squared w.r.t Constant 0.4729Adjusted Rho-Squared w.r.t Constant 0.4699Number of Observations 2874

* Significant at 10%. ** Significant at 5%. *** Significant at 1%

System ID Simulation Theory Hardware Discussion 19

• Negative constant suggests that direct pattern is generally preferred when no other information is available.

• Consistent with the other model, direct tours are found to have longer dwell time than the peddling.

• The chance of choosing peddling decreases if the minimum distance between stops or the average truck load per leg increases.

• Peddling patterns are more likely to be associated with semi trucks or tours that visit a distribution center, a retail store, or an external location, involve miscellaneous or transportation goods, or travel to a TAZ with high household density.

Summary Findings of Individual Tour Model

System ID Simulation Theory Hardware Discussion 20

• Tour chaining models are able to capture the interrelated nature of individual tours in a vehicle’s daily operation

• Individual tour model does not capture multiple-tour or mixed strategies, which are in many logistics are preferred strategies

• Tour chaining model gives much more detailed insight to choice preference–External location is shown positively related to peddling in the

individual tour based model and furthermore positively related to single peddling and mixed patterns in the tour chaining model

– Shipments of miscellaneous products are likely to use peddling strategies but less likely so in a mixed strategy as suggested by the daily tour chaining model

– Visits to distribution centers are not only strongly correlated with peddling strategies but more so with multiple peddling.

Individual Tour-based versus Daily Tour Chaining Approach

System ID Simulation Theory Hardware Discussion 21

• This study has presented an investigation of tour chaining based versus single tour based approach to understanding the complex process of urban commercial vehicle movements.

• The advantage of tour chaining model has been proven by comparing the individual tour and daily tour chaining models.

• Tour chaining reflects the interrelated decision process of linked trips.

• With mostly shipment related explanatory variables, this model would be useful in predicting future commercial vehicle daily tour patterns choice according to the changes of shipment demand in urban areas.

Conclusions

System ID Simulation Theory Hardware Discussion 22

• There are several limitations associated with this study that require further research. –Due to confidentiality, many critical variables (e.g., size, revenue,

etc.) were not available to the study. –Only the alternative specific variables were included in the model,

and the absence of generic variables has restricted the model explanatory capability.

–Decisions of daily tour chaining vary by shipping companies. No information about shippers was available.

– Some of the real-world constraints and conditions were not included in the model: driver work hour regulations, availability and locations of rest stop in relation to the destinations for examples, are not being considered in the model.

– Since the study dataset is composed of samples from 5 different study areas, spatial variations were not considered due to simplicity.

Conclusions (cont’d)

System ID Simulation Theory Hardware Discussion 23

• This study is funded by the National Center for Freight and Infrastructure Research and Education (CFIRE) at University of Wisconsin, Madison.

• We thank TxDOT for providing the data and generous support to the subject matter.

Acknowledgements

System ID Simulation Theory Hardware Discussion

Thank you!