Προηγμένες τεχνικές παράλληλου προγραμματισμού και πλέγματος για συστήματα ασύρματων επικοινωνιών

Abstract

The goal of this doctoral thesis is to deal with the increased computational requirements of specific problems concerning the simulation, optimization and planning of modern wireless and mobile communication systems, by employing parallel and distributed processing techniques and computational grid technologies. While the use of advanced and increasingly complex techniques and models (for example smart antennas, MIMO techniques, stochastic optimization methods, deterministic propagation prediction) becomes more frequent for wireless communication systems, the computational requirements of the concerning problems increase rapidly. The computation of these problems may become very time consuming or even practically infeasible in certain conditions. In the scope of the thesis, three specific problems regarding simulation, optimization and planning of wireless and mobile communication systems have been studied and the development and implementation of suitable methods for dealing with their increased computational needs has been attempted. A parallel radio-wave propagation prediction model based on ray tracing techniques and the electromagnetic theory of images has been developed. The implementation has been based on the computational paradigm of the message passing interface. Experimental results from utilizing the parallel model for problems of different sizes are presented, which demonstrate that the parallel algorithm achieves almost ideal scalability in all cases. Also, a distributed particle swarm optimization platform has been developed, based on the computational paradigm of mobile software agents. The particle swarm optimization algorithm entails a large number of computations of a cost function, which in the case of wireless communication systems can become time and memory consuming. The approach adopted enables the parallel and distributed processing of the algorithm on a number of heterogeneous computational nodes connected to each other via a network, providing ease and flexibility at deployment, as well as easy extension. Experimental results are presented by exploiting the system for the solution of an antenna array radiation pattern optimization. Finally, a networked environment for the simulation of wireless communication systems in a grid infrastructure has been developed, based on web technologies. The environment provides a safe web portal based on portlet specifications as the interface to the grid middleware, which hides from the end user the complexity of using the grid middleware. The simulation environment has been employed for executing a number of simulations and statistics from the execution of several WCDMA system simulation scenarios on the grid are presented. This doctoral dissertation has been supported by PENED 2003 project. The project is cofunded by the European Union (80%) and national resources (20%).