A computationally efficient genetic algorithm for MIMO broadcast scheduling

In conventional single-input single-output (SISO) systems, the capacity is limited as base station can provide service to only one user at any instant. However, multiuser (MU) multiple-input multiple-output (MIMO) systems deliver optimum system capacity by providing service to multiple users (as many as transmit antennas) simultaneously according to dirty paper coding (DPC) scheme. However, DPC is an exhaustive search algorithm (ESA) where the user encoding sequence is important to transmit data to multiple users. Exhaustive search becomes imperative as the search space grows with number of users and number of transmit antennas in the MU MIMO system. This can be treated as an optimization problem of maximizing the achievable system sum-rate. In this paper, it has been demonstrated that combined user and antenna scheduling (CUAS) with binary genetic algorithm (BGA) adopting elitism and adaptive mutation (AM) achieves about 97–99% of system sum-rate obtained by ESA (DPC) with significantly reduced computational and time complexity. It has been shown that BGA is able to find the globally optimum solution for MU MIMO systems well within the time interval of modern wireless packet data communications. However, it is interesting to observe that BGA is able to find a solution to CUAS close to the optimum value quite rapidly. In this paper, it is also shown that BGA with elitism and AM achieves higher throughput than limited feedback scheduling schemes as well.