Improving Spectral Efficiency of MIMO Ad Hoc Network via Greedy MCS Packing

This paper presents a solution for improving the network spectral efficiency (NSE) of a MIMO ad hoc network by simultaneously increasing the number of concurrent links in the network while maximizing the spectral efficiency of each. Our approach combines multiple physical layer techniques and balances the trade-offs of them. Assuming only the channel state information to the intended receiver, a two-level iterative algorithm is designed and presented. The inner loop uses an algorithm, which we refer to as the "greedy MCS packing"(GMP), generates the generalized (eigen-) beam forming, performs power allocation and chooses proper modulation and coding scheme (MCS). The GMP attempts to maximize the rate while packing them in as few eigen-channels as possible. The outer loop is used for frequency sub-band selection. It uses a heuristic algorithm to choose the number of "spectral segments" used by each TX-RX pair to further reduce overall network interference. Simulation results show that our algorithm yields as much as 71% improvement over a related previous work, which also combines multiple MIMO techniques and considers finite MCS rate with imperfect channel information. We further investigate the improvements in network spectral efficiency (NSE) when our baseline GMP approach is augmented by nonlinear Successive Interference Cancellation (SIC) at the receiver. While the NSE gain brought by this SIC-enhanced receiver is quite limited, our simulation shows that more concurrent links can be supported compared with GMP scheme using an MMSE receiver.