Algorithm-Architecture Co-Optimization of Bit-Loading Algorithm for Delay Constrained Adaptive Wireless Systems

Demands for higher throughput and performance along with increasing system and terminal constraints ( bandwidth, battery, regulatory spectral-masks etc. ) are pushing not only for the usage of efficient algorithms with minimal complexity but also for adaptive on-the-need basis usage of system resources to meet the time-varying channel constraints. adaptive modulation (Bit-Loading in discrete terms) has long been proposed as a technique to counter the time and frequency selective nature of the wireless system. For more than a decade, huge amount of research-work has been dedicated to improve the convergence time of discrete bit-loading algorithms for optimal allocation of bits to different sub-carriers of a multi-carrier system with respect to the channel response. However, no study as to our knowledge has been performed in quantifying the algorithm complexity in actual time for a given underlying architecture with respect to the coherence-time constraints of a real channel scenario. In this paper, we first explore the relationship between the coherence-time constraint of a real wireless system scenario (WiMax channel with varying Doppler) and the quantified complexity in time for different bit-loading algorithms with different underlying architecture profiles. We also propose a genetic algorithm based methodology for the selection of the optimum (on-the-need basis) architecture profile that meets the system constraints.