AI-assisted telecommunications network management

The authors describe an AI-assisted, real-time, centralized network management prototype that consists of two cooperating AI (artificial intelligence) components: the LARS (learning and recognition systems) and a rule-based expert system (RBES). The LARS system uses neural networks for the detection and isolation of communications network anomalies. This offers significant advantages over expert systems an conventional algorithms when dealing with complex patterns, ill-defined problems, and noisy input. The two-layer multiple manifold architecture greatly facilitates the training of the LARS system. The RBES is a real-time, data-driven expert system that is activated by the arrival of diagnostic messages from the LARS system. It uses diverse types of knowledge and reasoning techniques to assess the situation and recommend the application or removal of appropriate routing and traffic flow controls. A discrete-event network simulator (NETSIM) is assisting the development and testing of the above prototype,. The network management prototype can handle single and multiple anomalies, including node and link failures, link degradation, link congestion, and general overload