A prediction algorithm for feedback control models with long delays

It is well known that feedback control may be difficult in a network with long propagation delays (with large bandwidth-delay products). Various feedback-based congestion control algorithms considered for networks with long delays have been reviewed from control theoretical viewpoints. Most existing control algorithms are shown to provide a threshold-based control technique. However, as the propagation delay becomes significant, the past history over the entire time lag may have to be incorporated into the control strategy. An adaptive algorithm based on the optimal least mean square error technique is presented. The algorithm attempts to predict a future buffer size from the weight (slope) adaptation of unknown functions which are used for feedback control. Simulation results are presented, which show the effectiveness of the algorithm