Optimization of TCP Algorithm for Wired–Wireless Channels Based on Connection State Estimation

Abstract

The paper is devoted to the enhancement of the TCP data transmission through a heterogeneous (wired – wireless) communication channel. The idea is to include the current channel state estimate into the congestion control algorithm. The estimate is based on the statistical data available on the sender node: the packet acknowledgments, losses and time-outs flows. The corresponding mathematical model describes the channel state as a finite-state Markov jump process and the observable flows of losses and time-outs as Cox processes with state-dependent intensities. The high-frequency flow of packet acknowledgments is effectively represented by its diffusion approximation. The optimal filtering problem for the controllable observation system is properly stated and solved. The obtained high-precision state estimates are further incorporated into the modified TCP congestion control algorithm. The performance of the proposed modification is demonstrated in comparison with the existing versions of TCP.