Professional Documents
Culture Documents
196
explicit logistics choices is presented. In section three, the current The next layer is a disaggregate logistics model that forecasts
framework of FAME model is explained and some results of logistics decisions such as shipment size and mode choice at the
simulating freight flows in the United States using FAME are scale of firm-to-firm. The last layer is the network model that
presented. The forth section introduces the proposed FAME II assigns aggregate commodity flows to the traffic network.
framework. It also briefly discusses the methodology and data
requirements to operationalize the model. Then the last section In a recent study, Roorda et al. [11] proposed an inclusive agent-
presents the conclusions and the direction for future work. based freight micro-simulation framework. They discussed the
diversity of actors in freight system and explained the interactions
2. BACKGROUND between those actors in different markets. This study is still in
Tavasszy et al. [4] can be considered as pioneers in developing a progress and the framework has not been made operational.
freight modeling framework with logistics decisions. They However, the framework emphasizes on more complex supply
developed an aggregate model that includes logistics choices, chains and discusses new aspects of freight demand modelling
called the Strategic Model for Integrated Logistic Evaluation such as outsourcing of logistics services to a third party logistics
(SMILE), in the Netherlands. The theoretical bases of SMILE (3PL). They argue that making this model framework operational
were founded at the beginning of 90’s and the model was first is a controversial task and seems to be data intensive.
used in 1997. SMILE’s framework is based on a three-level chain While the majority of freight demand models that are
modeling approach that includes main freight activities: developed/proposed for the US in the literature work at the
production, inventory, and transportation. Logistics choices are regional scale and forecasts freight flows at aggregate zonal level,
simulated by repeating activities in the chain. Taking into account the goal of the current study is to significantly enhance those
the warehouse usage and determination of the optimum locations models by incorporating logistics decisions in a highly
of the warehouses were the main innovations of SMILE at that disaggregate level. On the other hand, the proposed modeling
time. In 2004, a program was implemented by the Dutch Ministry framework has a large geographical coverage and can simulate
of Transport, Waterways, and Public Works to improve the freight flows at the national scale and thus can be used for long
SMILE’s model structure, which resulted in proposing a new term policy analysis and planning. The model considers firms as
version of the model, called SMILE+. Bovenkerk et al. [5] the individual decision making units and estimates commodity
explains the improvements and calibrations resulted in SMILE+ in flows at the disaggregate scale of firm-to-firm.
detail.
In another effort, Boerkamps et al. [6] developed a disaggregate
commodity based freight transportation model, called GoodTRIP, 3. FAME
for the city of Groningen in the Netherlands. This urban freight 3.1 The framework
model replicated the supply chain patterns and urban truck tours,
This section provides a brief overview of FAME framework and
and thus depicted a picture of how the logistics decisions are
its models. Further elaboration of the framework and models can
made and affect urban commercial truck traffic. The model
be found in Samimi et. al. [12] for interested readers. The
framework consists of different actors and markets that function
framework has a modular structure which can be seen in Figure 1.
in the freight system while representing those actors’ interaction
Each of five modules consists of some models that develop some
through different markets in the system.
tasks. More detailed explanation of modules and their tasks are
GoodTRIP model structure was also used by Wisetjindawat and discussed in the following sections.
Sano [7], to develop a behavioral micro-simulation model for
urban freight transportation. This model includes three steps of the Firm Synthesis:
four steps in the traditional four-step modeling approach and Introducing individual decision-makers
incorporates behavior of decision makers. The model has been
used to forecast freight movements in Tokyo Metropolitan Area
and it has been validated with actual data. They used five percent
of the operating establishments in the study area for the simulation Supplier Selection:
purpose and provided the results including truck origin- Determining trade relationships/supply chains
destination matrices and approximate vehicle kilometer traveled
in the study area by each truck type [8].
In a more sophisticated study, De Jong and Ben-Akiva [9] Shipment Size Modeling:
developed a disaggregate logistics modeling framework which Using an iterative proportional fitting model
ideally works at the firm-to-firm level. They considered a logistics
cost function as their objective function in the framework. The
logistics decisions such as shipment size and mode choice are
determined by minimizing the objective logistics cost function. Mode Choice:
The paper discussed the model framework and the required data Modal split between truck and rail
for the model set-up for national freight movements in Norway
and Sweden. Later Ben-Akiva and de Jong [10] reintroduced their
freight model framework as the Aggregate-Disaggregate-
Aggregate (ADA) freight model system. Presumably, the ADA Network Analysis:
model system is a freight transportation model that can be used at Traffic assignment
the international, national or regional level. The ADA model
system includes three distinct layers. The first layer is an Figure 1. Fame operational framework
aggregate model that predicts production to consumption flows.
197
3.2 Data
3.1.1 Firm synthesis To develop the aforementioned models four category of data are
required such as, information on business establishments in the
As discussed earlier, the decision-making units in FAME model
first module, aggregate freight flows (OD matrix) in the second
are firms. However, since considering all existing firms in the
module, detailed information on a sample of individual shipments
U.S. in the simulation process can make some computational
and supply chains in the third and forth modules, and
burdens, some sort of aggregation is used to tackle this problem.
specifications of the transportation networks used in last module.
Thus, a new concept of firm-type is proposed. A firm-type is a
Except the information on a sample of individual shipments, all
group of firms with the same industry type and employee size in
other datasets are publicly available in the U.S. This cuts the data
the same geographic location in the zoning system. It is assumed
collection costs and is considered as the model strength. The
that all firms in a firm-type group have the same characteristics
information on the individual shipments are gathered through an
and similar behavior in freight decision-making process. In this
establishment survey. The survey was conducted during 2009-
module a list of all business establishment named County
2011 in three waves. The results of the first wave are used in
Business Patterns (CBP) [13] is used to generate firm-types with
development of current framework. Survey had three main parts
their basic characteristics. In total, 45,206 firm-types were
including, general information on establishments, information on
identified in the U.S.
individual shipments, and contact information [3]. In total, 966
3.1.2 Supplier selection establishments participated in the survey which resulted in 1840
In the second module, supplier selection, trade relationships are individual shipments. Respondents from a diverse range of
formed and firm-types are connected to make supplier-buyer industry type participated in this survey. Thus, various commodity
pairs. Then total annual zone-to-zone commodity flows are types are covered in the survey. Also detailed information about
allocated between supplier-buyer pairs. It makes sense to model the establishments and their logistics decisions are obtained from
supplier selection along with the logistics decisions such as the survey that makes the dataset a unique source of information
shipment size, etc. However, in the FAME model, supplier to develop detailed logistics models.
selection is modeled separately and before modeling the logistics
choices. This decision was considered primarily to counteract data 3.3 Simulation results
limitations and to avoid making the logistics strategies more In total, 45,207 firm-types are synthesized for the entire U.S. with
sophisticated. In this module, supplier choice is determined for a high level of industrial sector detail, and across several
each firm-type using a bi-level fuzzy rule based model that is employment categories which generate around 13.47 billion tons
developed based on firms’ characteristics. As suggested by commodity flows. 40 classes of commodities based on Standard
Stadtler [14], the financial status and location of the firms are Classification of Transported Goods (SCTG) are considered in
considered among the most significant characteristics in the this simulation. Some of the major results of simulation are
supplier selection decision and therefore are used in the FAME’s presented in this section. Since there is no disaggregate freight
supplier selection model. data available for the public use, the simulation results are
aggregated and compared with the publicly available aggregate
3.1.3 Shipment size datasets such as, Freight Analysis Framework data (FAF) [17] and
Now that the annual commodity flows between each supplier- Commodity Flow Survey (CFS) [15].
buyer pair are known, the characteristics of shipments can be Figure 2 presents the results of the supplier selection module
identified. In the third module, an Iterative Proportional Fitting where trade relationships are formed and commodity flows
(IPF) model is used to determine the shipment size of commodity between firm-types are allocated. The results are compared with
flows between each supplier-buyer pair. The observed shipment FAF data which shows a satisfactory match between the FAME’s
size distribution by the traveled distance in the U.S. [15] is used to simulated flows and FAF’s flows.
break down the annual commodity flows into three clusters: small
(less than 1,000 lb), medium (1,000-50,000 lb), and large (more
than 50,000 lb).
198
As it can be seen in Figure 3, less than three percent of shipments
Waste/scrap
Misc. mfg. prods. are transported in small size and the large shipments have the
Furniture
FAME (KTon) highest share. However, two different distributions will be
Precision instruments obtained if the distributions of shipment size by different
Transport equip. FAF (KTon)
Motorized vehicles
commodities are examined (see Figure 4). From 40 categories of
Electronics commodity groups, shipment weight distribution of 11
Machinery commodities follow a monotonically increasing form as shown in
Articles-base metal Figure 4a, while other 29 commodities have a distribution more or
Base metals
Nonmetal min. prods.
less like Figure 4b.
Textiles/leather
Printed prods.
Paper articles
Newsprint/paper
12
Wood prods.
Logs 10
Plastics/rubber
8
% of Shipments
Chemical prods.
Fertilizers
Pharmaceuticals 6
Basic chemicals
Coal-n.e.c.
Fuel oils
4
Gasoline
Crude petroleum 2
Coal
Metallic ores 0
Nonmetallic minerals
Gravel 0 2 4 6 8 10
Natural sands Shipment weight
Building stone
Tobacco prods.
Alcoholic beverages
“a”
Other foodstuffs
Milled grain prods. 12
Meat/seafood
Animal feed
Other ag prods.
10
% of shipments
Cereal grains
8
0
1,000,000
1,250,000
1,500,000
1,750,000
2,000,000
2,250,000
2,500,000
250,000
500,000
750,000
6
4
Figure 2. Total annual commodity flows
2
40%
compared with publicly available data in Figures 5 and 6. As
30% figures illustrate, FAME results are pretty close to FAF data and
have a satisfactory match with CFS data. According to the FAME
20% model 85 % of shipments by weight are transported with truck
and the rest with rail. In addition to the tonnage split by mode,
10% value and ton-mile of commodities for each mode were obtained
and shown in the figures.
0%
1-999 1,000-10,000 >10,000
Shipment weight (KTON)
199
100%
of consolidation, distribution, and transshipment centers. In this
module, the process of shipping chain formation for each pair of
90% buyer-supplier and for each commodity will be determined. From
80% a transportation point of view, the shipping chain can be defined
70% as the linkage between supplier and buyer of goods. It is worth
60% noting that in business and marketing, the term distribution
channel is used by researchers with a similar definition [6 and 7].
50% CFS 2007 A shipping chain can be a combination of one or more links. A
40% FAF3 link is the connection between two points inside the shipping
30% chain. For example, a link can connect the supplier to the
FAME distribution center, warehouse, or any intermediate handling
20%
facility or can connect the supplier to the buyer directly. Figure 7
10%
illustrates some of possible shipping chains between each pair of
0% supplier-buyer (i.e., firms in the FAME model).
Tonnage Ton-mile Value
Figure 5. Relative share of truck shipments
link Destination
Origin Firmi
Firmj
60%
link link Destination
Origin Firmi
Firmj
50%
link link Destination
Origin Firmi
40% Firmj
CFS 2007
30% link link Destination
FAF3 Origin Firmi
Firmj
20% FAME
link link link Destination
Origin Firmi
Firmj
10%
: Intermodal terminal : Distribution center : Consolidation center
0%
Tonnage Ton-mile Value Figure 7. Examples of Possible Shipping Chains
200
logistics decisions for a business could be very high, it is should be also determined in a way that minimizes the overall
envisioned that the primary method of modeling logistics choices logistics costs. It should be noted that all three logistics choices
in this layer would be to minimize total logistics costs. In this are decided at the disaggregate level of firm-to-firm. As a result,
case, the logistics cost could be considered as the combination of decision in each task of the logistics chain may affect decisions in
inventory and transportation cost [10]. other two tasks, for example, shipment size and transportation
mode choice would affect logistics costs and consequently have
impacts on shipping chain selection. Therefore, it is crucial to
consider these three modules in the logistics chain as interrelated
Production-Consumption chain factors that collectively define the optimum choices.
It is also worth noting that the logistics model proposed for
Firm Generation FAME II differs from the former framework used in FAME
Introducing individual decision-makers with their model. In the earlier framework, logistics decisions are treated
characteristics and geographical distribution
separately and shipment sizes are determined by using the existing
distributions based observed data (such as observed distributions
from the US Commodity Flow Survey). Such a process works like
a descriptive model that tries to replicate the existing trend. The
Supplier selection new model attempts to build a causal model which is policy–
Determining trade relationships between sensitive and has the potential to assess various freight strategies
firm-types and logistics policies.
The last layer, transportation chain, deals with the network
analysis. In this step, freight flows are assigned to the network in
order to assess the freight transportation impacts on the traffic
condition, environment, safety, etc. As noted before, network
Logistics chain
condition can significantly affects the decision-making process in
the logistics chain and supplier selection modules. In the current
operational version of FAME model the framework with no
Shipping chains formation interrelation between logistics choices and network analysis is
utilized. It is critical to consider the network conditions in the
overall decision-making process. Hence, the results of the network
analysis model should be captured in the previous tasks by using
recursive loops in the framework.
201
participants were asked to form the exact shipping chain of their Analysis, International Transportation Operation Research 5
shipments and provide detailed information such as the type of (6), 447-459.
intermediate handling facilities used, haul time and transportation [5] Bovenkerk, M. 2005. SMILE+, The New and Improved
cost to each facility, waiting time at the facility, mode of Dutch National Freight Model System. In: European
transportation used, etc. The survey also included some new Transport Conference, Strasbourg. 3-5 October 2005.
questions on the supplier selection preferences and criteria,
backhauling, percentage of empty truck trips, and truck ownership [6] Boerkamps, J.H.K., van Binsbergen, A.J., and Bovy, P.H.L.
by firms. 2000. Modeling Behavioral Aspects of Urban Freight
Movement in Supply Chains, Transportation Research
5. CONCLUSION Record, 1725, 17-25.
A nationwide freight micro-simulation model name FAME was [7] Wisetjindawat, W., and Sano, K. 2003. A Behavioral
developed using several public data sources and results of a Modeling in Micro-simulation for Urban Freight
nationwide establishment survey. The model works at Transportation, Journal of the Eastern Asia Society for
disaggregate level of firm-to-firm and shows an acceptable match Transportation Studies, 5, 2193–2208.
in the validation process with the available data. However, some
simplifications have been made in developing logistics models [8] Wisetjindawat, W., Sano, K., Matsumoto, S., and
and it lacks some critical logistics choices such as, use of Raothanachonkun, P. 2007. Microsimulation Model for
intermediate handling facilities. The current study has proposed a Modeling Freight Agents Interactions in Urban Freight
manageable modeling framework that incorporates detailed Movement. In: Proceedings of 86th Annual Meeting of the
logistics elements to the operational freight transportation micro- Transportation Research Board, Washington DC, 21–25
simulation model, FAME. Several new elements are included in January 2007.
the framework to more precisely predict the commodity flows. In [9] de Jong, G., and Ben-Akiva, M. 2007. A Micro-simulation
particular, the new logistics module, called shipping chain Model of Shipment Size and Transport Chain Choice,
formation, focuses on the use of different types of intermediate Transportation Research Part B, 41, 950–965.
handling facilities between producer and consumer. [10] Ben-Akiva, M., and. de Jong, G. 2008. The Aggregate-
Estimation of the new logistics module requires information on Disaggregate-Aggregate (ADA) Freight Model System, In:
individual shipments’ characteristics such as origin and M.E. Ben-Akiva, H. Meersman and E. van de Voorde eds.
destination, shipping chain, value of the goods, modes, shipment Recent Developments in Transport Modelling: Lessons for
size, or shipment frequency, use of consolidation centers, The Freight Secto', Emerald press, 2008, 117-134.
distribution centers, use of intermodal terminal facilities, among
other factors. It also needs spatial information on all types of [11] Roorda, M.J., Cavalcante, R., McCabe, S., and Kwan, H.
intermediate handling facilities, inventory costs, transport costs, 2010. A Conceptual Framework for agent-based modeling of
and all other related costs to the logistics costs. In order to obtain logistics services. Transportation Research Part E: Logistics
the required information to develop the models, a large scale and Transportation Review, 46 (1), 18-31.
survey was conducted at the national level that provided the [12] Samimi, A., Mohammadian, A. and Kawamura, K. 2010. A
required data to estimate the proposed models. Estimating the Behavioral Freight Movement Microsimulation Model:
required models and developing the micro-simulation tool is Method and Data, Journal of Transportation Letters: The
currently under process. International Journal of Transportation Research, 2 (1), 53-
62.
6. ACKNOWLEDGMENTS
The authors appreciate the National Center for Freight, [13] U.S. Census Bureau: Economic Planning and Coordination
Infrastructure, Research and Education (CFIRE) at the University Division, 2008. County Business Patterns. Available at:
of Wisconsin-Madison and Illinois Department of Transportation http://www.census.gov/epcd/cbp/view/cbpview.html,
(IDOT) for funding this study. All responsibility for the contents accessed February 2011.
of the paper lies with the authors. [14] Stadtler, H. 2005. Supply chain management and advanced
planning - basics, overview and challenges. European
7. REFERENCES Journal of Operational Research, 163, 575-588.
[1] Chow, J.Y.J., Yang, C.H., and Regan, A.C. 2010. State-of-
[15] Bureau of Transportation Statistics, Commodity Flow Survey
the Art of Freight Forecast Modeling: Lessons Learned and
(CFS) Preliminary Tables. United States, 2008. Available at:
The Road Ahead. Transportation 37 (6), 1011-1030.
http://www.bts.gov/publications/commodity_flow_survey/pr
[2] Regan, A.C., and Garrido, R.A. 2001. Modeling freight eliminary_tables_december_2008/index.html, accessed
demand and shipper behavior: state of the art, future February, 2011.
directions. In:. Hensher, D eds, Travel Behavior Research:
[16] Samimi, A., Pourabdollahi, Z., Mohammadian, A. and
The Leading Edge, Pergamon, Amsterdam, 2001, 1–30.
Kawamura, K. 2012. An Activity-based Freight Mode
[3] Samimi, A., Mohammadian A., and Kawamura, K. 2010. An Choice Microsimulation Model. In: Proceedings of the 91th
Online Freight Shipment Survey in the United States: Annual Meeting of the Transportation Research Board,
Lessons Learned and Nonresponse Bias Analysis, In Washington, D.C., January 2012.
Proceedings of the 89th Annual Meeting of the
[17] U.S. Department of Transportation, Federal Highway
Transportation Research Board, Washington, D.C., January
Administration. 2010. Freight Analysis Framework (FAF3)
11-15, 2010.
Data and Documentation: 2007-2040. Available at
[4] Tavasszy, L.A., Smeenk,B., and Ruijgrok, C.J. 1998. A DSS http://www.ops.fhwa.dot.gov/freight/freight_analysis/faf/.
For Modeling Logistics Chains in Freight Transport Policy
202
[18] McFadden, D., Winston, C., and Boersh-supan, A. 1985. [19] Vernimmen, B., and Witlox, F. 2003. The Inventory-
Joint Estimation of Freight Transportation Decisions under Theoretic Approach to Modal Choice in Freight Transport:
Nonrandom Sampling, In: Daughety, A.F. eds. Analytical Literature Review and Case Study. Brussels Economic
Studies in Transport Economics, Cambridge University Review, 46 (2), 5-29.
Press, 1985, 137-157.
203