Adaptive Reduced-Rank MIMO Decoder for Military Communications Patrick J Honan Ufuk Tureli Zhongren Cao Digital Design Solution, Inc. Dept. of Electrical Engineering & California Institute for Telecommunications & Clifton, NJ 07013 Computer Engineering Information Technology email: honan@onebox.com Stevens Institute of Technology University of California, San Diego Hoboken, NJ 07030 email: zcao@soe.ucsd.edu email: utureli@stevens.edu Abstract— We propose an adaptive reduced-rank solution for decoding spatial multiplexed multiple-input multiple-output (MIMO)wireless communications based on the multi-stage Wiener filter (MSWF). The huge spectral efficiency gains promised by spatial multiplexing schemes, such as vertical Bell Labs space time (V-BLAST) and space-division multiple access (SDMA), are difficult to realize for less than ideal channel conditions. This is particularly true for military applications where extreme channel conditions such as co-channel interference and intentional jamming can be expected. We show that the MSWF, first proposed by Goldstein and Reed for sensor array application, as combined here with spatial multiplexing decoding method successive interference (SIC) leads to significant signal subspace compression or rank-reduction. We propose for the MSWF-SIC a novel rank-reduction stopping rule that adapts to channel correlation and interference conditions. As a result our solution converges much faster than full rank methods permitting shorter training length and robustness to co-channel interfer- ence and jamming. Also, significant computational reduction is realized making this method a good canidae for portable applications where computational resources are limited. Bit error rate (BER) versus SNR for the correlated channel and for co- channel interference and SINR versus the MSWF rank-reduction and length of training interval are presented. I. Introduction During the last decade, extensive research has been carried out for the multiple-input and multiple-output (MIMO) an- tenna array systems [1]. It is recognized that MIMO system can provide high spectral efficiency and/or improve the re- liability for wireless communications. Two major techniques, space-time coding and spatial multiplexing [2], have been pro- posed for the MIMO wireless communication systems. Space- time coding exploits the diversity provided by the MIMO system to improve the reliability of the wireless communi- cation. Spatial multiplexing exploits the capacity provided by the MIMO system to support high-speed communication. Most theoretical MIMO studies so far use an idealized MIMO channel model and assume independent and identically dis- tributed fading among different transmit-receive antenna pairs. However, spatial fading correlation of the MIMO channel is a major consideration in designing practical MIMO systems [3]. Spatial multiplexing is realized by transmitting independent data signals from individual antennas. The receiver exploits the rich multi-path between the transmit and the receive antenna arrays to separate different signal stream, yielding a linear increase in capacity. A layered space-time architecture, named as vertical Bell Laboratories Layered Space-Time (V-BLAST), has been proposed in [4] that reaps the spatial multiplexing gain in the rich-scattering environment. Multiplexing gains degrade quickly in the presence of spatial correlation [5]. MIMO multiplexing systems are of particular interest to the military wireless communications. In typical modern urban digital battlefields and sensor networks, the communication between the airborne support, such as a helicopter, and the battle unit on the ground is of great importance to the tactical operations. A possible urban battle scenario with airborne support is depicted in Figure 1. It can be envisioned that MIMO multiplexing systems are applicable to military wireless communication systems in a variety of diversified scenarios. Also, the existence of co-channel interference or intentional jamming should be expected. Therefore, military MIMO multiplexing receivers are required to be robust to co- channel interference or jamming, as well as adaptive to various degrees of channel correlation. In the literature, the V-BLAST decoder technique of nulling/cancelation based on either zero forcing (ZF) and minimum mean square error (MMSE) have been proposed [4]. ZF decoder is particularly sensitive to spatial correlation [5]. Linear full-rank MMSE filter based receiver [6], [7] is more ro- bust to channel correlation, but performance quickly degrades in the presence of spatially correlated co-channel interference as shown in [8], since sufficient estimation of interference covariance matrix or details of spatial structure would require unacceptably long training intervals. Furthermore, complexity of this method is prohibitively high for portable and mobile applications with limited computational power resources. In this paper, a novel adaptive reduced-rank MIMO receiver structure based on the multi-stage Wiener filter (MSWF) [9] is proposed. Similar to MMSE based receivers, training periods are required to estimate the filter coefficients. MSWF recur- sively compresses the observation vectors during training into a smaller subspace without estimating the covariance matrix. The rank (subspace dimension) needed to achieve or exceed MMSE performance can be much less than the dimension of the signal subspace. Therefore, MSWF can be implemented in a reduced-rank fashion. Combining MSWF with successive interference cancelation (SIC) further reduces subspace order of the remaining transmit layers yet to be decoded. We propose for the MSWF-SIC a novel rank-reduction stopping mule solution that adapts to channel correlation and interference conditions. This adaptation rule is expected to address channel variability concerns of a military deployed MIMO commu- nications system. In addition, rank reduction brings several