ANN approach for congestion control in packet switch OBP satellite

The paper investigates the application of neural networks in a packet switch OBP satellite system to estimate the traffic intensity in the downlink queue and to predict the traffic load status. Two neural networks are used. The first one estimates the traffic intensity from the number of packets that arrive in a frame and the other calculates the congestion probability in the next two round trip delay. We show that in this case the linear estimator trained by a special signal outperforms the nonlinear one. Also, due to the long term dependency in the traffic and the validity of the Poisson model for a very short interval, the congestion probability formula is approximated with a neural network.