Intelligent systems for traffic flow management: A qualitative modeling approach
José Cuena · International Journal of Intelligent Systems · 1992
An approach for the design of intelligent systems for traffic control is presented. First, the problem characteristics are described. Then, the possible knowledge representation techniques are proposed. the types of knowledge to be represented are: — The data interpretation knowledge for problem identification based on the present and predicted situations. — The knowledge for proposing control decisions to solve present and predicted problems. — The knowledge for possible short-term situations prediction with the present control policy assumed as constant. The first two types can be represented by traditional techniques such as rules or frames. the predictive knowledge has to be defined on qualitative modeling techniques, an aspect of intensive research nowadays. Two types of predictive modeling techniques are suggested: (a) the qualitative simulation of flow along an axis which is the case of the flow in urban motorways where it is very important the preservation of an operational capacity. A confluence version is proposed for the partial derivative equations of unsteady traffic flow in a lane. (b) the qualitative modeling of the flow in networks where a qualitative version of the flow assignment to paths in the network is used. A proposal of knowledge representation for every prediction type is described. Although specific aspects of traffic control are presented the approach discussed can be considered as a general knowledge representation concept which combines traditional techniques and qualitative models in order to infer decision aspects for management purposes in a professional engineering domain. Finally, some knowledge structuring and acquisition aspects are considered.