A Swarm-Based Approach to Generate Challenging Mazes.

Joanna Kwiecień

doi:10.3390/e20100762

Abstract

Swarm intelligence draws its inspiration from the collective behaviour of many individual agents interacting with both one another and their environment. This paper presents a possibility to apply a swarm-based algorithm, modelled after the behaviour of individuals operating within a group where individuals move around in the manner intended to avoid mutual collisions, to create the most challenging maze developed on a board with determined dimensions. When solving such a problem, two complexity measures are used. Firstly, the complexity of the path was assumed to be a quality criterion, depending on the number of bends and the length of the path between two set points that was subjected to maximisation. Secondly, we focus on the well-known concept of the maze complexity given as the total complexity of the path and all branches. Owing to the uniqueness of the problem, consisting in the maze modification, a methodology was developed to make it possible for the individuals belonging to their population to make various types of movements, e.g., approach the best individual, within the range of visibility, or relocate randomly. The test results presented here indicate a potential prospect of application of the swarm-based methods to generate more and more challenging two-dimensional mazes.

Highlights

Based on the observations of the collective behaviour of the swarms of insects or herds of vertebrates, learning from each other and co-operating jointly for a common objective, a new field of computational intelligence was developed: swarm intelligence
The goal of this paper is to present a method of generating complex mazes, with the use of the existing swarm intelligence mechanisms
We carried out many experiments for different size of maze, we report our results for square mazes

Summary

Introduction

Based on the observations of the collective behaviour of the swarms of insects or herds of vertebrates, learning from each other and co-operating jointly for a common objective, a new field of computational intelligence was developed: swarm intelligence. The solutions occurring in natural systems, including the mechanisms that coordinate movement and work of a community, are applied in many real problems. It is used to resolve the issues relating to optimizations, finding best routes, assignments, developing work schedules, task arrangement or object grouping. The rules governing a swarm or herd were developed with time by instincts, and some of the rules were improved, owing to the individual agent’s ability to learn. A coordination system of a group of individuals, being a decentralised system, composed of autonomous units, is responsible for the organisation of tasks required for resolving a specific problem

Objectives

Methods

Results

Conclusion