Agent Swarm Optimization: a paradigm to tackle complex problems. Application to Water Distribution System Design

Agent Swarm Optimization (ASO) is a generalization of Particle Swarm Optimization (PSO) orientated towards distributed artificial intelligence, taking as a base the concept of multi-agent systems. It is aimed at supporting decision-making processes by solving either single or multi-objective optimization problems. ASO offers a common framework for the plurality of co-existent population-based algorithms and other heuristics. A particle from a PSO swarm, an ant from an ACO (Ant Colony Optimization) system, and a chromosome from a GA (Genetic Algorithm) structure do exhibit different behaviour. Yet, they all share a common feature: each represents a potential solution for the problem to be solved. In a combined environment, a PSO particle could help reinforce pheromone on the ants’ paths; an ant could be reproduced with a chromosome; a chromosome could be the leader of a particle swarm, and so on. This framework is a dynamic environment where new agents/swarms can be added in real time to contribute to the solution of the problem. During the solution process, the own user can add new agents/swarms to the environment and even contribute to the solution process with problem-based personal proposals. In this work the ASO framework is described, and used to solve a complex problem in water management, namely the optimal design of water distribution systems (including, sizing of components, reliability, renewal and rehabilitation strategies, etc.) using a multi-objective approach.