Applications of Particle Swarm Optimization to Digital Communication

Abstract

The whole class of evolutionary computing algorithms is inspired by the process of evolution in nature. Compared to the traditional optimization algorithms, a few striking features of these algorithms include their ability to address non-differentiable cost functions, robustness to the dynamically changing environment, and implementation on parallel machines. However, it was not until one and half decade ago, when these algorithms attracted researchers and got acknowledgement in terms of their application to the real world problems. The main reason behind this increased interest of the researchers owes to the ever increasing computing power. As a result evolutionary computing algorithms have been widely investigated and successfully applied for a number of problems belonging to diverse areas. In this dissertation the standard binary particle swarm optimization (PSO) and its soft version, namely soft PSO (SPSO) have been applied to four different problems of digital communication. Due to the exponentially growing computational complexity with the number of users in optimum maximum likelihood detector (OMLD), suboptimum techniques have received significant attention. We have proposed the SPSO for the multiuser detection (MUD) in synchronous as well as asynchronous multicarrier code division multiple access (MC- CDMA) systems. The performance of SPSO based MUD has been investigated to be near optimum, while its computational complexity is far less than OMLD. Particle swarm optimization (PSO) aided with radial basis functions (RBF) has been suggested to carry out multiuser detection (MUD) for synchronous direct sequence code division multiple access (DS-CDMA) systems. The MUD problem has been taken as a pattern classification problem and radial basis functions have been used due to their excellent performance for pattern classification. The two variants of PSO have also been used in a joint manner for the task of the channel and data estimation based on the maximum likelihood principle. The PSO algorithm works at two different levels. At the upper level the continuous PSO estimates the channel, while at the lower level, the soft PSO detects the data. The simulation results have proved to be better than that of joint Genetic algorithm and Viterbi algorithm (GAVA) approach.