Self-limitation, dynamic and flexible approaches for particle swarm optimisation

Swarm Intelligence (SI) is one of the prominent techniques employed to solve optimisation problems. It has been applied to problems pertaining to engineering, schedule, planning, networking and design. However, this technique has two main limitations. First, the SI technique may not be suitable for the online applications, as it does not have the same aspects of limitations as an online platform. Second, setting the parameter for SI techniques to produce the most promising outcome is challenging. Therefore, this research has been conducted to overcome these two limitations. Based on the literature, Particle Swarm Optimisation (PSO) was selected as the main SI for this research, due to its proven performances, abilities and simplicity. Five new techniques were created based on the PSO technique in order to address the two limitations. The first two techniques focused on the first limitation, while the other three techniques focused on the latter. Three main experiments (benchmark problems, engineering problems, path planning problems) were designed to assess the capabilities and performances of these five new techniques. These new techniques were also compared against several other well-established SI techniques such as the Genetic Algorithm (GA), Differential Equation (DE) and Cuckoo Search Algorithm (CSA). Potential Field (PF), Probabilistic Road Map (PRM), Rapidly-explore Random Tree (RRT) and Dijkstra’s Algorithm (DA) were also included in the path planning problem in order to compare these new techniques’ performances against Classical methods of path planning. Results showed all five introduced techniques managed to outperform or at least perform as good as well-established techniques in all three experiments