Cross-Layer Design for TCP Throughput Optimization in Cooperative Relaying Networks

In this paper, we investigate the transmission control protocol (TCP) throughput in cooperative relaying networks and take an cross-layer design approach when selecting a relay to optimize the TCP throughput. A first-order finite-sate Markov channel (FSMC) is used to model the wireless time varying channels, and the TCP throughput is estimated as a function of physical layer signal-to-noise ratio (SNR) and link-layer frame size and retransmission times. Since relay selection is crucial in improving the TCP performance, we proposed a stochastic decision making approach to select the optimal relay for every TCP packet according to the states of each relay. We formulated the cross-layer TCP throughput optimization problem as a restless bandit system and obtained the statistically optimal relay selection policy, which has an indexability property and can be easily implemented in real system. We compare the proposed scheme through simulations under different parameters of physical layer and link-layer, simulation results show that the TCP throughput can be improved significantly by the optimal relay selection scheme.