Cross-layer throughput optimization for delay and QoS constrained applications

In this article we recast the problem of throughput optimization in cross-layer architecture for delay sensitive applications. Unified mathematical framework was developed for maximizing user throughput in generalised wireless environment. The optimization variables adapted include symbol rate (at the application layer), packet length (at the MAC layer), constellation size (at the physical layer) and number of retransmissions of M-PSK digital modulation schemes. We derive optimization equations for each of the above degrees of freedom in closed-form which maximizes the throughput subject to a prescribed packet loss probability constraint. Interior-point optimization algorithm was used to solve the resultant constrained non-linear optimization problem. Numerical results reveal that no successful packet decoding is possible at very low average signal-to-noise (SNR) regime as the packet loss constraint is not met when channel condition is very bad irrespective of adaption. At moderately low SNR, it is sufficient to adapt packet length while keeping symbol rate fixed at the maximum value while joint optimization of both the number of retransmissions and the constellation size parameters while keeping the packet length at an appropriate value is required in the high average SNR regime, to achieve maximum throughput over a myriad of wireless fading environments. Results also showed that the tighter the packet loss constraint, the lower the throughput becomes.