Regional Freight Transport Distribution Models

Today freight transport covers great importance in the field of the transportation system and in general in the economic system. It represents one of the most important elements in the national economy development process. In spite of the importance of freight transport, only in the last ten years the research devotes itself to this field. The most investigate topics are the national freight transport (Cascetta E. et alii, 1996; Cascetta E. and Iannò D., 2000; Cascetta E., 2001; Nuzzolo A. and Russo F., 1997). Recently the research has focused its efforts on movement and management of urban freight transport and problems concerned with them (Ogden K.W., 1992; Oppenheim N., 1995; Russo F. and Comi A., 2002). From the available data published by transport statistics (e.g. White Book 2001 and CNT 1998) emerges the importance of trips related to the distance band 50-200 Km, which are generally the regional trips. This freight band has not been investigated in literature and for it are generally used aggregate non behavioural models. Moreover in this band operate different typologies of vehicles in relation of different types of trips. It could possible to identify two types of processes: one for heavy goods vehicles (HGVs) and another for light goods vehicles (LGHs). This distinction is necessary because generally the HGVs are solely used for the transport of the commodities and semimanufactured, while a large proportion of LGVs is used from Service Company. Furthermore, the path choice is related to different requirements (and so we need of different models to simulate them) and the available data to calibrate the models are found in a different way. In this paper is analysed the problem of the regional freight transport. Firstly, are analysed the different types of vehicles used in this band and are described the models proposed in literature for their analysis. Secondly, is presented a behavioural demand model and is proposed a possible structure for a calibration model from aggregate data and from traffic counts. This data can be combined in different ways, obtaining different estimators of unknown parameters of demand model, in relation to the classic or Bayesian interpretation of the initial estimation and to the assumption of the probability distribution of the measurement, assignment and model errors. Many estimators can be expressed in the general form as a constrained multi-dimensional optimisation problem: ( ) [ ] ) ( z ) ) ( ( z ) ( z argmin = * 3 1 S f̂ , b d M̂ d̂ , b d ˆ , b B b + + ∈ 2 β β (1) where zi(.) can be considered as a “distance” measure between the unknown vector and the estimated one; b is the unknown parameter vector of demand models; β̂ is the parameter vector estimation; d (β) is the demand function; d̂ is the demand vector estimation; M̂ is the assignment matrix estimation; f̂ is the vector of traffic counts; SB is the feasibility set of the parameter vector. An intuitive interpretation of the problem (1) is to search the vector β to be closest to the estimated vector β̂ , and, once assigned to the network, to produce flows closest to the counts f̂ ; at the same time the demand function d (β) should be closest to the estimated vector d̂ .