Algorithms for building annular structures with minimalist robots inspired by brood sorting in ant colonies

Wilson, Matt; Melhuish, Chris; Sendova-Franks, Ana B.; Scholes, Samuel

Abstract

This study shows that a task as complicated as multi-object 'ant-like annular sorting' can be accomplished with 'minimalist' solutions employing simple mechanisms and minimal hardware. It provides an alternative to 'patch sorting' for multi-object sorting. Three different mechanisms, based on hypotheses about the behaviour of Leptothorax ants are investigated and comparisons are made. Mechanism I employs a simple clustering algorithm, with objects of different sizes. The mechanism explores the idea that it is the size difference of the object that promotes segregation. Mechanism II is an extension to our earlier two-object segregation mechanism. We test the ability of this mechanism to segregate an increased number of object types. Mechanism III uses a combined leaky integrator, which allows a greater segregation of object types while retaining the compactness of the structure. Its performance is improved by optimizing the mechanism's parameters using a genetic algorithm. We compare the three mechanisms in terms of sorting performance. Comparisons between the results of these sorting mechanisms and the behaviour of ants should facilitate further insights into both biological and robotic research and make a contribution to the further development of swarm robotics.

• Journal Article Type: Article
• Publication Date: Sep 1, 2004
• Journal: Autonomous Robots
• Print ISSN: 0929-5593
• Publisher: Springer (part of Springer Nature)
• Peer Reviewed: Peer Reviewed
• Volume: 17
• Issue: 2-3
• Pages: 115-136
• DOI: https://doi.org/10.1023/B%3AAURO.0000033969.52486.3d
• Keywords: collective behaviour transition, genetic algorithm, ant; sorting, swarm intelligence, swarm robotics
• Public URL: https://uwe-repository.worktribe.com/output/1036752
• Publisher URL: http://dx.doi.org/10.1023/B:AURO.0000033969.52486.3d
• Additional Information: Additional Information : The paper represented an original strand of research exploring behaviour transitions in a collective of minimalist robots. The work was bio-inspired and, as such, was undertaken by a multidisciplinary team of roboticists and biologists. The study provided a hypothesis of action selection transition in robots and a theoretical explanation along with detailed results with a fleet of robots which took careful observation over a number of weeks. This early work in bio-inspired collective systems has paved the way for BRL's current research programme in swarm systems. The paper is used in robotics degree courses and has 15 google scholar citations.