On the Convergence of Genetic Scheduling Algorithms for Downlink Transmission in Multi-User MIMO Systems

In a multi-user MIMO system using a successive precoding method such as dirty paper coding, it is combinatorially complex to determine the optimal set of users to schedule and the proper order to encode their signals in order to optimize a utility function in a scheduling algorithm. Genetic algorithms represent a fast suboptimal approach to reducing the complexity of the search. In this paper, we build upon prior work that implements scheduling via genetic algorithms. We examine the impact of parameter values within the adaptive mutation rate of the algorithm on its convergence time. We demonstrate that although there is a range of values for the parameters that yields similar near-minimum convergence times, it is nonetheless important to ensure that the parameters are tuned to be within that range. In one case, tuning the parameter values reduces the time of convergence to less than 30% compared to that achievable with the initial parameter values. We also demonstrate that the proper parameter values are dependent on both the number of transmit antennas and the number of users in the pool of users to be scheduled. A simple equation is proposed that is linear in the adaptive mutation parameters to tune the values for different numbers of transmit antennas and users.