Human learning principles inspired particle swarm optimization algorithms
Muhammad Rizwan Tanweer
Date of Issue2017-05-29
School of Computer Science and Engineering
Centre for Computational Intelligence
These days, the nature of global optimization problems, especially for engineering systems has become extremely complex. For these types of problems, nature inspired search based algorithms are providing much better solutions compared with other classical optimization methods. Among them, the Particle Swarm Optimization (PSO) algorithm has been mostly preferred due to its simplicity and ability to provide better solutions. PSO algorithm simulates the social behaviour of a bird swarm in search of food where the birds are modelled as particles. The limitations associated with PSO have been extensively studied and different modifications, variations and refinements to PSO have been proposed in the literature for enhancing its performance. The idea of utilizing intelligent swarms motivated towards exploring human cognitive learning principles for PSO. As discussed in learning psychology, human beings are known to be intelligent and have good social cognizance. Therefore, any optimization technique employing human-like learning strategies should prove to be more effective. This thesis addresses the use of human learning principles inspired strategies for the PSO algorithm. The major contributions of the thesis are: • Self-Regulating Particle Swarm Optimization (SRPSO) algorithm. • Dynamic Mentoring and Self-Regulation based Particle Swarm Optimization (DMeSRPSO) algorithm. • Directionally Driven Self-Regulating Particle Swarm Optimization (DD-SRPSO) algorithm. • Incorporation of a constraint handling mechanism in the structure of the DDSRPSO algorithm. The Self-Regulating Particle Swarm Optimization (SRPSO) algorithm is inspired from the human self-learning principles. SRPSO utilizes self-regulation and self-perception based learning strategies to achieve an enhanced exploration and a better exploitation. The self-regulated inertia weights are employed only for the best particle whereas all the other particles perform search employing self-perception of the global best search direction. The perception is dynamically changed in every iteration for intelligent exploitation. The effect of human learning strategies on the particles has been studied using CEC2005 benchmark problems and the performance has been compared with the state-of-the-art PSO variants. The results clearly indicate that SRPSO converge faster closer to the global optimum with a 95% confidence level. Further, human beings utilize multiple information processing strategies during the learning process and collaborate with each other for better decision making. Integration of socially shared information processing will further enhance the performance. Therefore, a new algorithm referred to as Dynamic Mentoring and Self-Regulation based Particle Swarm Optimization (DMeSR-PSO) algorithm has been proposed incorporating the concept of mentoring together with the self-regulation. Here, the particles are divided into three groups consisting of mentors, mentees and independent learners. The elite particles are grouped as mentors to guide the poorly performing particles of the mentees group. The independent learners perform search using self-perception based learning strategy of the SRPSO algorithm. Tested on both the unimodal and multimodal CEC2005 benchmark problems the DMeSR-PSO has shown improved convergence than the SRPSO algorithm. Further, the robustness of the algorithm has been tested on CEC2013 problems and eight real-world optimization problems from CEC2011. The results indicate that DMeSR-PSO is significantly better than other PSO variants and other population based optimization algorithms with a 95% confidence level, yielding an effective optimization algorithm for real-world applications. Both SRPSO and DMeSR-PSO are rotationally variant algorithms and therefore the performances have not been significant on the rotated problems. To overcome this, a directionally updated and rotationally invariant SRPSO algorithm has also been developed named as Directionally Driven SRPSO (DD-SRPSO) algorithm. Here, the poorly performing particles are equipped with complete social perception guidance. Other particles are randomly selected to perform search either by using self-perception based learning strategy of SRPSO or by applying a rotation invariant strategy. The performance of DD-SRPSO tested on rotated problems from CEC2013 proves that DD-SRPSO is significantly better than SRPSO. Its performance, compared with other algorithms on CEC2013 benchmark problems clearly indicates that DD-SRPSO is significantly better than selected algorithms on a wide range of problems. Further, a new constraint handling mechanism has been incorporated in the DDSRPSO structure referred to as DD-SRPSO with constraint handing mechanism (DDSRPSOCHM). Next, the application of DD-SRPSO-CHM in optimizing multi-stage launch vehicle configuration has been studied. In a multi-stage launch vehicle configuration, the multiple objectives are converted into a single objective with constraints and these are efficiently handled by DD-SRPSO-CHM. Comparative analysis on the problem suggests that DD-SRPSO is converging faster towards the solution. By incorporating human-like behaviour in the PSO algorithm, the developed variants have shown a faster convergence closer to the optima over a diverse set of problems indicating that the algorithms are potential choice for complex real-world applications. In the future, the algorithm will be extended for solving multi-objective optimization problems. The equality constraint handling mechanism has already been implemented in the DD-SRPSO algorithm which can be further extended for the inequality constraints. Furthermore, more human learning strategies can be explored for performance enhancement.