One of the biggest obstacles for developing multi-agent freight system for the commodity distribution
part is to identify the connection between each pair of shipper and customer. At present, there are many
models developed using logit model for calculating the fractions. However, the question is to calibrate
that kind of model, one need to know exactly which the shipper for each customer. This is obviously
impossible to be obtained this detailed data in reality. The objective of this study is therefore to develop a
commodity distribution model for a multi-agent system to overcome the problem of data availability for
model calibration. The model is based on the decisions of each individual customer on purchasing
commodities from each individual shipper. The commodity distribution can be viewed as a supplier
location choice model to which a discrete choice model can be applied. The commodity flows between
shippers and customers are calculated based on a logit model, constrained by the amount of the
commodity generated and attracted by each of the companies considering the distance between companies
and attractive factors of the shippers. The model is developed and applied for the urban freight movement
in the Tokyo Metropolitan Area.