Scenario selection optimization in system engineering projects under uncertainty: A multi-objective Ant Colony method based on a learning mechanism

Abstract

This paper presents a multi-objective Ant Colony Optimization (MOACO) algorithm based on a learning mechanism (named MOACO-L) for the optimization of project scenario selection under uncertainty in a system engineering (SE) process. The objectives to minimize are the total cost of the project, its total duration and the global risk. Risk is considered as an uncertainty about task costs and task durations in the project graph. The learning mechanism aims to improve the MOACO algorithm for the selection of optimal project scenarios in a SE project by considering the uncertainties on the project objectives. The MOACO-L algorithm is then developed by taking into account ants' past experiences. The learning mechanism allows a better exploration of the search space and an improvement of the MOACO algorithm performance. To validate our approach, some experimental results are presented.

Towards an Integration of Systems Engineering and Project Management Processes for a Decision Aiding Purpose

“…A multi-objective Ant Colony Algorithm (MOACO) has been developed for this problem for its ability to solve such relevant combinatorial optimization problem in a reasonable amount of time. First results provided by this algorithm were presented in (Lachhab et al (2016)). Following on these works, an important improvement is to define a decision-aided tool, based on the optimization model, that integrates the standard industrial processes (the systems engineering process (SEBOK (2014)) and the project management one (PMBOK (2013))) in the early first phases.…”

“…The Optimization process includes a multi-objective ACO tool that provides a range of Pareto-optimal solutions and minimize the total cost, duration and risk of the SE project. The uncertainties about project goals (cost and duration) are modelled using single intervals (Lachhab et al (2016)). The lower bounds correspond to nominal values and the upper bounds to the maximum possible values (estimated).…”

“…At the end of the iterations, a Pareto-front of optimal scenarios is built in order to help a decision-maker to select one scenario that is about to be further developed and realized. In Lachhab et al (2016), first experiments were conducted and shown that the MOACO algorithm gives better results using a learning mechanism (denoted MOACO-L). The MOACO-L algorithm has given efficient solutions in terms of cost, duration and risk and has improved the mean performance of the MOACO algorithm with almost a difference of 8.84%.…”

A risk-based approach applied to system engineering projects: A new learning based multi-criteria decision support tool based on an Ant Colony Algorithm

“…The first results were presented in Lachhab et al (2016) where a simple instance of the problem was tested without any kind of framework to implement the tool. Thus, the aim of this section is to propose a global framework of integrated industrial processes where the optimization tool is used.…”