The use of prospect theory framework in constrained multi-objective particle swarm optimisation

Bunny, MN 2012, The use of prospect theory framework in constrained multi-objective particle swarm optimisation , PhD thesis, University of Salford.

[img] PDF
Restricted to Repository staff only until 31 July 2023.

Download (31MB) | Request a copy


Many practical problems in the real world nowadays can be formulated as constraint single or multiple objective optimisation problems with constraints. Particle swarm optimisation (PSO) is a population-based stochastic algorithm has been shown to be an effective optimisation method for solving these types of problems since it is capable of generating random multi-start points, it is simple to perform and it does not require gradient continuity. Despite the popularity of this approach, PSO still needs more adaptation to the guide selection mechanisms in order to improve the search capacity of the particles and achieve better convergence. Moreover, PSO lacks an explicit mechanism to handle In this work, new constrained PSO-based optimisation algorithms are proposed for solving both single and multi-objective optimisation problems. The proposed methods introduce new decision mechanism that inspired from human behaviour under risk into the guide selection of particles. This human behaviour was formalised by Kahneman and Tversky in 1979 into mathematical equations represented by prospect theory (PT). Including PT in the proposed methods help to direct the swarm towards the feasible region, encourage the swarm to explore a new area in the search space and consequently, improve the convergence to the optimal solution (or the Pareto- optimal in the case of multi- objective problems). The performance of the proposed methods are tested and evaluated firstly on constrained nonlinear single-objective optimisation problems (CNOPs) using sixteen well-known benchmark functions. Moreover, the proposed method are tested and evaluated secondly on constrained multi-objective problems (CMOPs) using fourteen benchmark problems and two mechanical design-engineering problems. The proposed methods are validated also by comparing them with the current established state-of-the-art algorithms in the area. The results showed that the proposed methods are competitive when compared to the other approaches and outperforms the other algorithms in many cases.

Item Type: Thesis (PhD)
Contributors: Nefti-Meziani, S (Supervisor)
Schools: Schools > School of Computing, Science and Engineering
Depositing User: Institutional Repository
Date Deposited: 04 Aug 2021 13:50
Last Modified: 04 Aug 2022 11:22

Actions (login required)

Edit record (repository staff only) Edit record (repository staff only)


Downloads per month over past year