Eigenmode Scheduling via Simulated Annealing for Multiuser MIMO Downlink with Successive Zero-Forcing Precoding

Abstract:
In this paper we consider the scheduling problem in multiuser multiple-input multiple-output (MIMO) systems with successive zero-forcing precoding and coordinated transmit-receive processing. We focus specifically on the problem of scheduling of downlink transmissions, while allowing a variable number of data streams per user. However, optimal scheduling employing exhaustive search for a constant number of streams per user is combinatorially complex and impractical, and this problem is further compounded by allowing a variable number of streams per user. To reduce the complexity we show that transmitting only on the dominant channel eigenmodes for each user achieves close to exhaustive search sum-rates, and consider a suboptimal metaheuristic scheduling algorithm based on simulated annealing. Simulated annealing approaches have been shown to provide quick and relatively good solutions to combinatorial optimization problems. We modify the simulated annealing approach by accounting for the impact of dynamic neighbourhood sizing on the convergence rates and overall performance of the algorithm. The performance of the simulated annealing scheduling is examined and compared to that of the exhaustive search. We demonstrate that our scheduler offers comparable sum-rate performance to the exhaustive search and that variable data-stream allocation offers significant performance gains over scheduling the maximal possible number of data streams per user.