Download files
Abstract
Evolving swarms have advantages over monolithic vehicles in a number of situations. They are able to distribute their sensors more widely, require less sophistication and are robust in that the failure of one element leads to the degradation of the swarm rather than complete failure. Significant benefits of these characteristics include greater endurance and the ability to operate in dangerous environments.
Two applications were developed in this thesis to investigate the technological feasibility of an ever evolving swarm in different technological areas. While one part, of the research, explored the ruggedness bestowed on a space system in a debris field by using distributed agents the other looked at how their performance could be improved in a much simpler scenario by using evolutionary algorithms. The aims of this thesis are to explore the feasibility of using an evolutionary method to develop the rules governing self-organised systems to coordinate and control swarms of satellites in debris avoidance manoeuvres and UAVs in marine search and rescue (SAR) missions. The proposed models are capable of improving the energy efficiency of two different methods by which a cluster of spacecraft can perform a scatter and re-gather manoeuvre to rapidly avoid collision with space debris; as well as allowing SAR co-ordinators to better manage their man-power and resources for an increased chance of target identification within a desired timeframe.
A swarm intelligent model was developed and implemented using the agent-based software, NetLogo© for SAR missions. The model used self-organised emergent behaviour with an evolutionary algorithm to optimise the effectiveness of the swarm in adapting to changing environments. The spacecraft debris avoidance case study model was developed in Matlab and validated by comparing it with a STK / Astrogator generated ephemeris 3D model within an orbital environment simulator. The elliptic formation validated through STK was able to achieve the same performance in Matlab in a continuous and nonlinear model through incorporating J2 perturbations.
The results show a swarm-intelligence based approach is effective, robust and scalable. Both of the developed scattering manoeuvres were able to maintain the desired cluster flight formation in J2 perturbations with Δv optimised to minimise fuel consumption. In addition, the SAR UAVs demonstrated the ability to form teams, fly in formation and find the targets with only initial human input required being to generate a fitness factor; with results showing that there is an effective division of labour.