Modelling Commercial Vehicle Empty Trips with Parameters That Depend on Trip Characteristics
J Holguin-Veras, J C Zorilla, E Thorson, Rensselaer Polytechnic Institute, US
freight demand modeling, empty trips, trip chains, commercial vehicles, intermodal transportation, transportation planning, transportation modeling
The 21st century has been characterized by the increasing role of information technology in everyday life. Modern computer systems allow for information to be transferred faster and safer through the Internet, thus making it convenient for consumers to shop online in the comfort of their homes. As this trend continues, the demand for lighter, higher-value goods increases, since they are likely to have a lower cost to the consumer on the Internet. This increase in demand, combined with population growth and other factors, is resulting in an increase in the amount of freight that has to be transported, particularly by truck.
A hindering factor to efficient ground transportation is the larger commercial vehicle traffic that increases congestion on the roads. This results in a significant increase in air and noise pollution as well as the number of dangerous traffic accidents involving trucks.
The increase in truck traffic, along with the accompanying increase in externalities, puts pressure on the trucking industry and on Metropolitan Planning Organizations (MPOs). The trucking industry will need to handle larger delivery volumes, facing lower revenues due to the level of competition and increasingly stringent regulations for externalities. MPOs will have to improve their planning processes in order to accommodate the ever increasing demand for goods transportation. In order to do this, the organizations need more efficient demand models than the ones currently in use. In many cases, MPOs use adaptations of passenger car models to estimate freight demand in the area. Although these simplistic approaches can sometimes provide rough estimates, they are fundamentally flawed since they do not capture the key dynamics of freight phenomena.
Currently there are two major platforms for modelling freight transportation demand: vehicle-trip and commodity based models (Ogden, 1978). Vehicle-trip models focus on modelling the actual number of vehicle trips, which has some practical advantages. Among them are the relative ease and high-quality with which data can be obtained due to an increasing number of Intelligent Transportation Systems. Also, since the model focuses on vehicle trips it has no problem generating the number of empty trips between regions. However, these models have two fundamental limitations. The first one is that these models cannot be applied to multimodal transportation because the vehicle-trip is in itself the result of a mode choice and the selection process is not represented in the data (Holgun-Veras, 2000, 2002). Furthermore, since the models assume that the vehicle is the unit of demand, as opposed to the commodity being transported, the model neglects the economic characteristics of the shipment that have been found to play a significant role in the majority of choice processes in the trucking industry (e.g., Holgun-Veras, 2002).
Commodity based models, as the name points out, focus on modelling the flow of goods from one region to the other (measured in a unit of weight). Since the commodities are the unit of demand, the modeler can capture the underlying factors that determine freight movement, such as value, weight, and volume. In this platform, the loaded trips are estimated by dividing the total flow from one region to the other by an average payload from all loaded trucks. The problem with commodity-based models is that they are unable to model empty trips, which can make up about 30 to 50 percent of the total trips in a region. This occurs since the commodity flow in one direction determines the loaded trips, but does not bear a relationship to the number of the empty trips in the same direction. To resolve this, some complementary models have been developed, such as Hautzinger's (1984), Noortman and van Es' (1978) and Holgun-Veras and Thorson (2003a). The complementary models developed by Noortman and van Es and Holgun-Veras and Thorson will be described and compared in this paper.
The empty trip models mentioned before are some of the approaches for modelling empty commercial vehicle traffic that are discussed in the paper, which also proposes some new empty trip models. The models range from simple nave formulations to some more complex ones involving trip chains, probabilities and memory components. The performance of the alternative formulations to model empty trips is assessed by applying these models to a sample data set from the Dominican Republic (see description in Holgun-Veras, 2003), as well as a data set from Guatemala City (see Holgun-Veras and Thorson, 2003a). The paper starts with some background information on the subject, followed by a brief description of previous developments in the area, a description of the test cases for the model, the methodology, and finally the results and conclusions.
This paper provides a comprehensive discussion of the theoretical developments pertaining to commercial vehicle empty trip models, the corresponding estimation procedures, empirical evidence, and practical implications and describes new mathematical formulations to model commercial vehicle empty trips. In doing so, the paper synthesizes and expands the state of the art of empty trip modelling. The paper introduces two novel formulations to estimate the probability of zero order trip chains that relax the assumption used by the authors in their previous works that this probability was constant and obtains optimal parameters for some of the models. The paper also discusses the empirical support for the fundamental assumptions used in almost all empty trip models. Overall, the new formulations provide a significant improvement with respect to the previous ones.
Association for European Transport