The DRIVE project ATTAIN (V1039) was started to evaluate the potential of Artificial Intelligence based software approaches dealing with traffic engineering problems which can be solved only inadequately by conventional methods or which have had to be excluded previously from computerization. The paper analyses the requirements which should be present in considered traffic problem areas to ensure a beneficial application of Artificial Intelligence to enhance conventional techniques or to tackle completely new domains. A decision tree has been set up which leads to the identification of classes of AI applications which are similar or analogous to a given TE problem. This classification is derived from classification is derived from generic problems to which AI is "typically" applied and of which practical experiences from related areas are available. Together with this a concise set of criteria and prerequisites concerning solution space, computational complexity, uncertainty of input parameters, presence of heuristics etc. is presented to support the traffic engineer in identifying problems which are suitable for applying AI. A major result of ATTAIN is the proposal of several prototypical projects which exhibit the best chances in proving Artificial Intelligence as a powerful tool for solving of up to the moment open questions or for enhancing of inadequately realized existing systems.
Abstract