|
|
MIT Electrical Engineering and Computer Science
EECS Event |
Tuesday, October 2, 2001
4:00 PM (reception following)
Room 35-225
LIDS Colloquium
Abstract
Wireless networks are unique among communication networks by virtue of the mobility and portability they enable for the end-user. The fundamental limitations on performance in wireless networks are wrought by physical propagation phenomena. A new paradigm that mitigates their adverse effects by employing multiple antennas at both the transmitter and the receiver is rapidly emerging as an acceptable practical solution. In theory, such multiple-input multiple-output (MIMO) systems provide enormous gains in data rate that are linear in the number of antennas. In practice, however, efficient signal design to achieve these rates is still an open problem. One system that is well studied is the coherent receiver, i.e., channel side information is fully available at the receiver and is unknown at the transmitter. A large variety of signal designs have been proposed for this scenario, running the gamut from designs that maximize signal reliability to designs that maximize information rate. Each of these designs is well understood in isolation, but to date there has been no cohesive analysis of general space-time codes that ties all these designs together.
In this talk we provide a unifying framework for the design and analysis of space-time codes that are linear in the input information symbols. A linear codeword is defined to be the sum of certain modulation matrices scaled by input symbols. Previous design of space-time codes is based only on minimizing the worst-case pairwise error probability (PEP), i.e., the probability of error between the closest pair of codewords. We take the entire constellation into account by employing the conventional union bound on probability of error. The union bound is in fact a tighter upper bound on the probability of error than the worst-case PEP. We derive necessary and sufficient conditions on a constrained set of modulation matrices that minimize the union bound. As special cases we obtain the well-known spatial multiplexing (SM) (or BLAST) scheme and orthogonal space-time block codes (O-STBCs). Of these, SM is known to be capacity-efficient, i.e., the code structure does not induce a loss in channel capacity. Orthogonal STBCs, however, incur a loss in capacity over channels with multiple receive antennas. We provide sufficient conditions on capacity-efficient codes for any number of receive antennas. Finally we design new linear space-time block codes based on both error probability and capacity that outperform previously proposed capacity-efficient codes.