Artificial Intelligence (AI) and, in particular, knowledge-based systems (KBS) are receiving increased attention within the transport engineering community. Several applications of such systems have been recently attempted in different areas and road traffic control is no exception. Objective of the DRIVE project V 1055 is developing and experimenting AI techniques in surveillance, control and management of traffic in signalised urban networks, and drawing indications on the benefits and the implications of such techniques for the next generation of RTI (Road Transport Informatics) systems. A system is being developed based on a blackboard architecture, where four knowledge-based (sub)systems represent specialised problem solving agents associated to different key aspects of traffic control operations, and a central blackboard provides a common representation of the traffic system and a data exchange facility for the four subsystems. The following problem areas are addressed by the prototype: (1) inference of additional traffic data from sensor supplied data, to enlarge the database available for control decisions; (2) analysis and interpretation of collected data to recognise traffic situations, detect critical events and make diagnosis; (3) qualitative prediction of traffic flow based on collected sensor data; (4) action planning and control decision making using, as a basis of the decision process, information about the level of performance of the current control strategy as well as a set of qualitative indicators of traffic behaviour in the network. The prototype is currently being developed with reference to a real case study, taking into account an area in the centre of Hamburg where a traffic actuated signal plan selection strategy is presently implemented. System prototyping and validation is carried out mainly using recorded historical data and a simulation model for integrated testing of the developed functions. A preliminary field test validation of the model is also planned in the case study area. The paper presents the overall conceptual design of the prototype and the various assumptions and AI methods used for realising the functions of each KBS. Details about the software tools used for system development are also provided.
Abstract