A new swarm optimal collective searching behaviour framework using decision-making under risk

Al-Dulaimy, AIA

Abstract

Swarm Intelligence (SI) is a recent computational intelligence technique which mimics
and makes use of the collective behaviour of flocks of birds or schools of fish for
solving search and optimization problems. There are several decision-making models
that have been introduced in the literature on collective searching behaviour. However,
those decision models are based on the Expected Utility Theory (BUT) and tend to
optimize the outcome value of the utility function; in other words, the logical decision
processes used in these models are rational and risk avert and they perform poorly
where risk is associated with the environment.
In this research, we will use the particle swarm metaphor as a model for the human
social group strategic adaptation for collective searching in a risky environment. The
objective is to show that endowing these particles with a human descriptive model
(irrational behaviour) from the field of psychology (using a theory named Prospect
Theory (PT)) can considerably improve the global searching ability of the swarm.
Unlike other proposed decision models, the BUT and other decision methods used in
collective searching, this proposed searching framework captures common human
decision-making attitudes towards risk, i.e., risk aversion and risk seeking, which is
vital for handling the risk of violating environmental constraints, hence improving the
exploration/exploitation during the evolutionary process.
The experimental results presented in this research provide evidence on the robustness,
the effectiveness and the practicability of the proposed framework when applied to
swarm robotics and many other engineering systems with a single objective function
under constraints.

• Thesis Type: Thesis
• Deposit Date: Jul 27, 2021
• Award Date: May 1, 2012