An experimental study on the robustness of integer-forcing linear receivers

Recent work has proposed the integer-forcing (IF) linear receiver architecture as a promising alternative to the joint maximum likelihood (ML) receiver. It has been proven that the IF linear receiver can operate very close to the (optimal) performance of the joint ML receiver, but with essentially the same implementation complexity as a zero-forcing (ZF) linear receiver. In this paper, we take the first steps towards a complete software-defined radio (SDR) implementation of the IF linear receiver. Using the Wireless Open-Access Radio Platform (WARP) and IEEE 802.11 protocols, we develop an OFDM-based experimental framework to evaluate the performance of IF linear receivers in realistic indoor settings, and compare it to the performance of ZF and joint ML receivers. Our framework includes a channel estimation protocol and algorithm for selecting the best integer matrix for approximating the channel matrix. We have performed indoor experiments for a 2 × 2 MIMO network, which demonstrate that the symbol error rate (SER) of the IF linear receiver indeed outperforms the ZF linear receiver and can operate close to the joint ML receiver. We also argue, via simulations, that IF continues to outperform conventional linear receivers, even in the presence of significant channel estimation errors.