In this paper, it was shown that empirical values of trip generations for modelling purposes can be obtained
not necessary from questionaires in transport companies. Of course, this gives general overview on truck usage,
but still does not give information about all trucks if is conducted on small area. Thus, wider truck survey like
TIUS (Truck Inventory and Usage Survey) made in USA should be conducted in Polish condition. This kind of
survey should answer questions about garaging places as well as daily truck trips. On the other hand, for
specified area, trip generations may be calculated using calibrated O-D matrix.
In model development different methods were used: trip generation rates, multiple regression and artificial
neural networks. Although the easiest to apply are the trip generation rates, it may not reflect commune
characteristic. Often in data bases only few independent variables characterizing commune are available. In this case trip generation rates are very suitable. Multiple regression and ANN are more complex and depending on the
number of explanatory variables may give better results. However in model verification, especially for Set B,
ANN and multiple regression models resulted with higher errors.