ACTIVE QUEUE MANAGEMENT (AQM) AND ADAPTIVE NEURO FUZZY INFERENCE SYSTEM (ANFIS) AS INTRANET TRAFFIC CONTROL

Congestion (traffic congestion) is one problem that requires continuous development of control system, including the application of artificial intelligence. This study proposes intranet traffic congestion control using algorithm of Active Queue Management (AQM) and Adaptive Neuro Fuzzy Inference System (ANFIS). The control improved router’s ability in determining congestion level through marking probability, enabling packet receiver to identify traffic load. Computer network representing intranet and traffic generation was formed based on simulation using Borland C++ Builder. The average of queue length (first variable: average) was simulated between 0-750 packets with minimum limit of 250 packets. Comparison between current and previous packet lengths (second variable: change) was calculated between (-200) and 200 packets. ANFIS was designed, trained, and tested using Mat lab, consisting of 2 inputs (average and change), and was achieved from reply packet as the result of traffic generation while the output produced was probability. Probability was suitable with certain congestion levels (low, medium, and high). Results of simulation showed that with the probability, traffic generation reduced transmission that, in turn, affected lower and delayed queue length. Transmission with low congestion showed queue length of 20.512%, medium congestion reduced queue length to an average of 44.971%, and high congestion reduced queue length up to 48.52%. Reduced queue length shortened delay to an average of 0.00445 seconds.