A Platform for Autonomous Drone Swarms
MetadataShow full item record
Search- and rescue operations requires participation from a lot of personnel, ranging from volunteers to non-governmental organizations and government agencies. Employing autonomous drone swarms for search can be useful in order to reduce search time by covering larger areas simultaneously than what human search crews are able to. In addition to reduce search time, drones may reduce both cost of the operation and risk of the search crew if they are operating in a hazardous search area, such as at sea or in mountainous terrain during bad weather. In cases where lives are at stake reduced search time correlates positively with increased chance of survival. This thesis presents the development of a virtual prototyping platform that facilitates development of autonomous drone swarms, and a guidance system built using this platform. The platform is designed to be modular in order to ensure adaptability with respect to how drones are configured for any given operation. Modularity is achieved by using a modified approach to agent-based design where logic is injected to the agent prior to execution. Map data from OpenStreetMap are used for both guidance and visualization. A map renderer for OpenStreetMap data has been created to facilitate interactivity and allows operators to specify the routes for which drones should travel. Hiking routes, motorways and places of interest is extracted from the map data and given to the drones, which is the basis for where the guidance systems lead the respective drones. Distributing drones in the specified search area has been solved as a task allocation problem, and three approaches to allocation has been tested; auctioning of tasks, allocation by genetic algorithms and lastly allocation by turn-based cherry-picking, where drones choose one task at a time until every task is chosen.