Flocking for multi agent system : split and merge algorithm

Abstract

The final year project is the extension of existing project with code name FAME.

FAME is C# language game engine software application based on the open source

rendering engine Orge3D developed to study and simulate steering behaviors of

multi autonomous agents.

Before this FYP, several steering behaviors has been studied and implemented

including arrival, goal seeking, wandering and flocking, the most interesting

behavior. Flocking algorithm attempts to simulate the beautiful natural phenomenon

of flocks of thousand birds, schools of countless fish or great herds of animals (In

computer science, ―this gathering of mass individuals‖ are commonly termed

flocking).Nonetheless, the problem of obstacle avoidance for flock of agents has not

been studied before. In this project, the problem of a flock obstacle avoidance is

analyzed. Inspired by the nature of flocking avoidance behavior, the obstacle

avoidance should comprise of ways to split the whole flock to steer around the

obstacle and merge back into the old flock afterward. This report is going to explain

in the details algorithm developed during 8 months and show that it is able to

produce realistic animation and deliver strong performance.

In addition, in FAME, a steering agent is a combination of several steering behaviors,

i.e, arrival behavior, goal seeking behavior, wandering behavior and flocking

behavior. Adding one more obstacle avoidance behavior increases chances to cause

conflicts and the optimization of steering behaviors has also not been thoroughly

investigated before. Thus, the problem of steering behaviors optimization and

solution are also discussed in this report.