Individual robots, while providing the opportunity to develop a bespoke and specialised system, suffer in terms of performance when it comes to executing a large number of concurrent tasks. In some cases it is possible to drastically increase the speed of task execution by adding more agents to a system, however this comes at a cost. By mass producing relatively simple robots, costs can be kept low while still gaining the benefit of large scale multi-tasking. This approach sits at the core of swarm robotics.
Robot swarms excel in tasks that rely heavily on their ability to multi-task, rather than applications that require bespoke actuation. Swarm suited tasks include: exploration, transportation or operation in dangerous environments. Swarms are particularly suited to hazardous environments due to the inherent expendability that comes with having multiple, decentralised agents. However, due to the variance in the environments a swarm may explore and their need to remain decentralised, a level of adaptability is required of them that can't be provided before a task begins. Methods of novel hormone-inspired robotic control are proposed in this thesis, offering solutions to these problems. These hormone inspired systems, or virtual hormones, provide an on-line method for adaptation that operates while a task is executed. These virtual hormones respond to environmental interactions. Then, through a mixture of decay and stimulant, provide values that grant contextually relevant information to individual robots. These values can then be used in decision making regarding parameters and behavioural changes.
The hormone inspired systems presented in this thesis are found to be effective in mid-task adaptation, allowing robots to improve their effectiveness with minimal user interaction. It is also found that it is possible to deploy amalgamations of multiple hormone systems, controlling robots at multiple levels, enabling swarms to achieve strong, energy-efficient, performance.
