A Distributed Intelligence Approach to Using Collaborating Unmanned Aerial Vehicles for Oil Spill Mapping
From swarming locusts to schools of fish, the complex emergent behaviors exhibited by multi-agent swarm systems in nature present a compelling basis for their application towards real-world challenges. This paper capitalizes on this potential by proposing a swarm-intelligence inspired approach towards mapping complex offshore oil spills — one that uses a collaborating team of small (inexpensive) unmanned aerial vehicles. By leveraging the idea of occupancy grids, a new probability map concept is developed to enable agent-level situational awareness, while significantly reducing computing overheads (image data to intelligence generation in <1 sec) and communication overheads (< 1.7 KB of average data sharing across the swarm agents). The probability map is further exploited for waypoint planning using the principles of swarm dynamics and a rule-based reasoning approach to allow for dynamic preference shifts towards map exploitation and exploration. Detection of oil is performed by using a generalizable concept of anomaly detection that is derived from a color-based segmentation approach. Two simulated case studies, derived from actual oil spill images, are presented with results highlighting the strengths of the proposed approach.