This paper presents an adaptive robotic swarm of Unmanned Aerial Vehicles (UAVs) enabling communications between separated non-swarm devices. The swarm nodes utilise machine learning and hyper-heuristic rule evolution to enable each swarm member to act appropriately for the given environment. The contribution of the machine learning is verified with an exploration of swarms with and without this module. The exploration finds that in challenging environments the learning greatly improves the swarm's ability to complete the task. The swarm evolution process of this study is found to successfully create different data transfer methods depending on the separation of non-swarm devices and the communication range of the swarm members. This paper also explores the resilience of the swarm to agent loss, and the scalability of the swarm in a range of environment sizes. In regard to resilience, the swarm is capable of recovering from agent loss and is found to have improved evolution. In regard to scalability, the swarm is observed to have no upper limit to the number of agents deployed in an environment. However, the size of the environment is seen to be a limit for optimal swarm performance.
|Number of pages||23|
|Journal||Australian Journal of Telecommunications and the Digital Economy|
|Publication status||Published - Jun 2018|
- Hyper Heuristics
- Machine Learning
- Mobile ad hoc networks
- Robotic Swarms