I. Introduction
Quickly responding to customer service requirement is focus of the current urban distribution service systems [1]. But compared with the general service system, urban distribution service system has the following features: Decision-making of the urban distribution program is closely related to the environment of distribution line, the spatial of distribution vehicles and so on, the current distribution dependent on the task environment; Urban distribution shows a cyclical repetition with time changing. Urban distribution is always in a fixed distribution network, so urban distribution line program has the typical repeatability. These features made the urban distribution program have a strong distribution repeatability, so if we can identify task environment of urban distribution cases, and build matching model supporting one to one mapping between the current urban distribution task and the historical urban distribution cases, will significantly improve reusability of mass historical distribution knowledge in the urban distribution system, to achieve responding quickly to customer requirements with urban distribution system.