Parallel Implementation of the Particle Swarm Optimization Algorithm on a Multiprocessor Embedded System with Network-on-Chip

Resumo

In recent years, with technological advancements, the need to solve complex problems has emerged in various areas of knowledge, such as data mining, combinatorial optimization, power systems, signal processing, pattern recognition, machine learning, and robotics. The key characteristic of these problems is their computational intensity, particularly in terms of execution time. In order to accelerate the problem-solving process, bio-inspired
algorithms have been developed, which aim to simulate the behavior found in biological systems, such as living organisms and ecosystems, to efficiently solve complex problems. Examples of these algorithms include Particle
Swarm Optimization, Ant Colony Optimization, Artificial Bee Colony, and Cuckoo Search. This work aims to obtain a parallel implementation of the Particle Swarm Optimization algorithm using a Multiprocessor Embedded System with Network-on-Chip. The parallelization strategies we employ are based on the Parallel Particle Swarm Optimization and Cooperative Parallel Particle Swarm Optimization algorithms, using master-slave, ring, and 2D
grid topologies. Based on the execution time obtained by each parallel algorithm and each employed topology
during the simulations, it will be possible to identify which parallelization strategy provides the best performance,
as well as the number of processors required. Currently, the results, when compared to the serial version of the
Particle Swarm Optimization algorithm, are promising.