In our daily life we are faced with all kinds of decision problems, whether we realize it or not, and many of these decision problems involve optimizations. Such problems are also common in all fields of applications. Although there are a rich set of tools available for the various classes of optimization problems, there is yet a continuous effort to improve the existing tools and to come up with new ones that can tackle computationally hard problems. This dissertation investigates a relatively recent nature inspired global optimization algorithm called Particle Swarm Optimization (PSO). PSO has been demonstrated to be an effective tool for a variety of optimization problems. Like other similar meta-heuristics, PSO has a few parameters that need to be properly tuned for each problem to achieve good performance. The first goal of this work is to present an automatic parameter optimization technique for the PSO algorithm to achieve enhanced performance.

Although PSO is shown to be efficient compared to other contemporary population based optimization techniques, for many continuous multimodal and multidimensional problems, it still suffers from performance loss when it is targeted onto embedded application platforms. Examples of such target applications could include small mobile robots, distributed sensor nodes in sensor network applications for environmental monitoring or for search and rescue applications. Implementation technologies in these and other similar application scenarios are commonly targeted on low footprint processors and programmable logic devices. The second goal of this dissertation is to develop a high-performance, efficient, flexible, modular and reusable hardware architecture for the PSO algorithm. The accelerated execution performance of the proposed architecture is demonstrated on standard mathematical benchmark functions as well as on large scale real world problem scenarios. A parallelization scheme for further speedup of the execution performance of the PSO algorithm is also presented and evaluated.