In cooperative communication single antenna mobiles can share their antennas to improve communication links and increase channel capacity. Node cooperation generates a virtual multiple-antenna system which attains close to optimal multiple input multiple output (MIMO) gains. At the same time, power line communication (PLC) offers stable high data rate for indoor communications, and also benefits from cooperative communication scheme since power lines can be used between relays. Thus it is of interest to investigate a hybrid PLC/wireless relying network. Towards that goal, in this paper, we derive analytic forms of single-user and multiuser MIMO channel capacities. We then evaluate relay channel capacities for two relevant cases. One when the channel is perfectly known to the transmitter, i.e. full channel state information (CSI) is known, and the other when the channel state is unknown (no CSI). We propose adaptive communication scheme which involves the decode-and-forward (DF) relaying and orthogonal frequency division multiplexing (OFDM). We investigate optimal allocation of transmit powers for the scheme at both source and relay nodes under total power constraints. For a realistic scenario our numerical tests indicate that PLC/wireless link performs better then a typical wireless link.

Summary form only given. Fully-adaptive space-time adaptive processing (STAP) is an optimal but impractical radar processing algorithm due to the overwhelming number of operations required. Hence, many different heuristic approaches are sought to approximate the optimal method with fewer number of operations. We introduce a framework for prototyping various STAP methods on a parallel system: describing STAP methods, building execution time models for basic building blocks, and using these models to automatically optimize performance of the algorithms. We also present the performance results from a COTS PC cluster.

Space-time adaptive processing (STAP) refers to a class of methods for detecting targets using an array of sensors. Various STAP methods use similar operations on different data or in different orders. We have developed a portable, parallel library of subroutines for prototyping STAP methods. The subroutines work on the IBM SP2 and the Intel Paragon under three different operating systems and three different communication libraries, and can also be configured for other systems. We provide execution-time models for predicting the performance of each subroutine. Using the library routines, we created a parallel version of element-space pre-Doppler processing, three parallel versions of higher-order post-Doppler processing, and two versions of PRI-staggered post-Doppler processing. We implemented a fourth version of higher-order post-Doppler processing, the hybrid method, which uses a combination of fine-grain and coarse-grain parallelism to reduce execution time. The hybrid method can be used to improve performance when a large number of processors is available. Our execution time models generally predict the best method and predict execution times to within 10 percent or better for large test cases.

We discuss implementations of block algorithms for recursive least squares (RLS for short) problems on ring distributed-memory multiprocessors. We consider the sliding rectangular window case which involves triangularization followed by updating and downdating of the data matrix. We compare several schemes for computing the current least-squares solution, including a direct back-substitution scheme and a scheme where the previous solution vector is updated to the current solution vector by adding the so-called Kalman gain vector. The techniques are implemented on a linear array of transputers and on the Intel iPSC/2 hypercube, and evaluated with respect to their execution time and numerical accuracy.

In this paper we discuss a Schur-type algorithm for solving indefinite Hankel systems of linear equations. The algorithm is derived using the language of linear algebra.

In this paper, we propose a new algorithm for computing a singular value decomposition of a product of three matrices. We show that our algorithm is numerically desirable in that all relevant residual elements will be numerically small In diesem Beitrag wird ein neuer Algorithmus zur Berechnung der Singulärwertzerlegung des Produktes dreier Matrizen vorgestellt. Es wird gezeigt, daβ der Algorithmus ein gutartiges numerisches Verhalten aufweist, weil alle relevanten Residualgröβen klein werden. Nous proposons dans cet article un nouvel algorithme de calcul de la décomposition en valeurs singulières d'un produit de trois matrices. Nous montrons que notre algorithme est attrayant du point de vue numérique du fait que tous les résidus pertinents sont numériquement petits.

In this paper we survey some parallel matrix algorithms that are used in signal processing. Our model of computation is restricted to the class of array processors.

We present a linearly connected array of O(n) cells that solves the linear least-squares problem for an (m + 1) × (n + 1) Toeplitz matrix in time O(m + n). The total storage required is O(n) words, i.e., only a constant per cell. The parallel algorithm described in this paper is based on the sequential QR factorization algorithm for Toeplitz matrices recently developed by the authors.

We present a new algorithm and systolic array for adaptive beamforming. Our approach improves on McWhirter's pioneering work in two respects. First, our algorithm uses only orthogonal transformations and thus should have better numerical properties. Second, the algorithm can be implemented on one single p × p triangular array of programmable processors that offers a throughput of one residual element per cycle.

The authors present novel fault-tolerance schemes for 2-D systolic implementation of a recursive least squares minimization problem with applications to beamforming problems. They show that the errors can be detected by examining only a few scalars. In the case of the transient errors the technique is self-correcting. The technique can be implemented with negligible algorithm modification and little additional hardware. The simplicity of the method invites its use in future systolic arrays.>

The authors introduce a modification of the hyperbolic Householder scheme that is demonstrably stable theoretically (according to an established stability criterion) and exhibits superior numerical behavior in simulations. The modified transform scheme effects downdating of Cholesky factors by applying conventional orthonormal, rather than hyperbolic, Householder transformations to the data. The latter have preferable numerical properties. However, the construction of these orthonormal operators itself requires hyperbolic computations. Thus the method is, in some sense, half hyperbolic and half orthonormal. There is no computational penalty incurred with these stabilized hyperbolic Householder transforms; they enjoy an operation count identical to that of their conventional counterparts. This approach offers roughly a twofold saving over schemes based on the multiple application of single vector downdating schemes, and it is just as stable.>

A modification of Newton's method for the solution of the equation F(x) ffi 0 on a multiprocessor computer is studied.A class of asynchronous Newton methods is introduced and an optimal method in this class, as well as its optimal parallel implementation, is shown.Then the optimal asynchronous parallel method is compared with the optimal asynchronous sequential method.It turns out that no matter how many processes are used, a Newton process (in the class of asynchronous Newton methods) can be speeded up by at most a factor of 4.