Using Parallel Computing for Adaptive Beamforming Applications

Recently, smart antenna systems have been widely considered to provide interfer- ence reduction and improve the capacity, data rates, and performance of wireless mobile commu- nication. Smart antenna arrays with adaptive beamforming capability are very efiective in the suppression of interference and multipath signals. The techniques of placing nulls in the antenna patterns to suppress interference and maximizing their gain in the direction of desired signal have received considerable attention in the past and are still of great interest using evolutionary algorithms such as genetic algorithms (GA) and particle swarm optimization (PSO) algorithm. In this paper, for adaptive arrays using space division multiple access (SDMA), the optimal ra- diation pattern design of smart antennas is developed based on the particle swarm optimization (PSO) technique. The PSO is applied to a 24-element uniform circular array (UCA) to calculate the complex excitations, amplitudes and phases of the adaptive array elements. The antenna elements consist of vertical (z-directed) half-wave dipole elements equally spaced in the x-y plane along a circular ring, where the distance between adjacent elements is dc = 0:5‚. It is found that the resulting beampattern optimized by the PSO required a large processing time which is not acceptable for an on line applications. Hence, the demand for a parallel solution that accel- erates these computations is considered. Therefore, a parallel version of PSO is proposed and implemented using Compute Unifled Device Architecture (CUDA) then applied on a graphics processing unit (GPU). The comparison is presented to show how the parallel version of the PSO outperforms the sequential one, thus an online procedure is available for time-critical applications of the adaptive beamforming. Over the last decade, wireless technology has grown at a formidable rate, thereby creating new and improved services at lower costs. This has resulted in an increase in airtime usage and in the number of subscribers. The most practical solution to this problem is to use spatial processing. Andrew Viterbi, founder of Qualcomm Inc., clearly stated: \Spatial processing remains as the most promising, if not the last frontier, in the evolution of multiple access systems". Spatial processing is the central idea of adaptive antennas or smart-antenna systems. Although it might seem that adaptive antennas have been recently discovered, they date back to World War II with the conventional Bartlett beamformer. It is only of today's advancement in powerful low-cost digital signal processors, general purpose processors (and ASICs | Application-Speciflc Integrated Circuits), as well as innovative software-based signal-processing techniques (algorithms), that smart antenna systems have received enormous interest worldwide. In fact, many overviews and tutorials have emerged, and a great deal of research is being done on the adaptive and direction-of-arrival (DOA) algorithms for smart-antenna systems. As the number of users and the demand for wireless services increases at an exponential rate, the need for wider coverage area and higher transmission quality rises. Smart-antenna systems provide a solution to this problem (3). The increasing interest of researchers in using low cost GPUs for applications requiring intensive parallel computing is due to the ability of these devices to solve parallelizable problems much faster than traditional sequential processors. The flrst applications of evolutionary algorithms (EAs) on GPUs have been developed to solve speciflc Adaptive beamforming problems; This paper presents an approach for the implementation of PSO algorithms on GPUs which, using the NVIDIA CUDA environment in the adaptive beamforming applications (1). For each running swarm a thread block is scheduled with a number of threads equal to the number of particles in the swarm The rest of the paper is structured as follows: Section 2 presents the problem formulation. Section 3 describes the experimental results. Finally, conclusions and remarks are presented in Section 4.