Design and Modelling of Decentralised Task Allocation Mechanisms in Groups of Mobile Agents

Division of labour is a fundamental field of research within the context of multi-agent (particularly swarm based systems) and multi-robot systems. Eusocial insects, for instance ants and bees, are known to display remarkable capabilities of allocating tasks to nest mates when the colony gets perturbed by any internal and/or external factors. Proper understanding of the underlying mechanisms of division of labour among these social insects would enable more effective designing and developing of artificial swarm based systems which in turn can be used in tackling various real world problems. At the same time, a properly built model can be used to serve as a platform for the biologists to test their research hypotheses. These key benefits have been the prime motivations of this thesis. The thesis is based on the behaviour of ant colonies and especially on how they allocate tasks in different situations. The objectives of the thesis are twofold: (1) to develop an artificial simulated system that is ant-like and (2) to explore, identify, develop and analyse task allocation strategies within the realms of colony performance.

The first objective of the thesis is approached by investigating the behaviour of ant colonies from the existing literature and modelling their behaviours using an agent based modelling approach. To determine whether the model has met the first objective, three questions are posed: (A) Is the emergent system scalable? (B) Is the emergent system flexible? and (C) Is the system robust? For a system to be ant-like, the system has to not only give the appearance of ant-like behaviour but also has to meet these three criteria. As a part of the second objective of the thesis, three task allocation strategies based on ant colony behaviour are proposed. Furthermore, the strategies are critically analysed to investigate the benefits of each of the strategies and also to discover under what circumstances which strategies would perform better. The research reported in this thesis is intended to provide a better understanding of the design issues of task allocation strategies thus enabling researchers to use this as a guide to design effective task allocation strategies within the concerned multi-agent systems.