Frequency allocation for green multiuser OFDM systems using evolutionary algorithm

Evolutionary approaches are usually shunned by engineers in real time applications because of high computational complexity. This paper makes a convincing argument that in computationally challenging problems like energy efficiency (EE) maximizing channel allocation, genetic algorithm (GA) is well worth considering. A two-step solution to the problem of finding the subchannel and power allocation that maximizes the EE of the OFDMA transmissions, under minimum rate and total power constraints, is presented. GA is used for subchannel allocation and is followed by optimal power allocation obtained via analytical methods. The fitness function necessary for the GA is the maximum EE for a fixed channel allocation, and is computationally expensive to be of any use here. Two different closed form approximations to the maximum EE, obtained from our previous work, are used as fitness functions. A new measure of computational complexity for evolutionary algorithms is introduced. Simulation results are used to show that while GA has comparable complexity to the best EE maximizing channel allocation protocol in the literature, it produces nearly double EE.