TY - JOUR
T1 - Distributed operation of collaborating unmanned aerial vehicles for time-sensitive oil spill mapping
AU - Odonkor, Philip
AU - Ball, Zachary
AU - Chowdhury, Souma
N1 - Publisher Copyright:
© 2019 Elsevier B.V.
PY - 2019/5
Y1 - 2019/5
N2 - Multiple simple agents working together to achieve a common complex goal embodies the underlying theme of swarm concepts, with decentralized decision-making serving as the new frontier for tackling challenges associated with scalability, fault tolerance, and communication constraints. This paper builds on this emerging paradigm to develop a distributed approach (called PSOil) for off-shore oil spill mapping using a team of unmanned aerial vehicles or UAVs. In-flight waypoint planning is achieved via a new particle swarm mechanics-inspired technique, employing a novel combination of anomaly detection for knowledge extraction, and a stochastic occupancy grid approach for timely processing and frugal sharing of knowledge (with net communications <1.7 KB/UAV every 10 waypoints). A total of ten real-world oil spill images, encapsulating complexities such as non-convex arbitrary shapes and disjointed segments, are studied in this work. Overall, the algorithm registered 55–90% completeness in mapping oil-covered areas. PSOil required around one-third the time necessary for an exhaustive survey and was found to be superior compared to a typical random walk and a spiral search approach w.r.t. mapping performance and efficiency, respectively. Further tests simulating increasing UAV team sizes (to deal with larger search areas) and the random loss of team members respectively illustrate the scalability and fault-tolerance characteristics of PSOil.
AB - Multiple simple agents working together to achieve a common complex goal embodies the underlying theme of swarm concepts, with decentralized decision-making serving as the new frontier for tackling challenges associated with scalability, fault tolerance, and communication constraints. This paper builds on this emerging paradigm to develop a distributed approach (called PSOil) for off-shore oil spill mapping using a team of unmanned aerial vehicles or UAVs. In-flight waypoint planning is achieved via a new particle swarm mechanics-inspired technique, employing a novel combination of anomaly detection for knowledge extraction, and a stochastic occupancy grid approach for timely processing and frugal sharing of knowledge (with net communications <1.7 KB/UAV every 10 waypoints). A total of ten real-world oil spill images, encapsulating complexities such as non-convex arbitrary shapes and disjointed segments, are studied in this work. Overall, the algorithm registered 55–90% completeness in mapping oil-covered areas. PSOil required around one-third the time necessary for an exhaustive survey and was found to be superior compared to a typical random walk and a spiral search approach w.r.t. mapping performance and efficiency, respectively. Further tests simulating increasing UAV team sizes (to deal with larger search areas) and the random loss of team members respectively illustrate the scalability and fault-tolerance characteristics of PSOil.
KW - Anomaly detection
KW - Distributed decision-making
KW - Oil spill mapping
KW - Swarm intelligence
KW - Unmanned aerial vehicles (UAV)
UR - http://www.scopus.com/inward/record.url?scp=85061701288&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85061701288&partnerID=8YFLogxK
U2 - 10.1016/j.swevo.2019.01.005
DO - 10.1016/j.swevo.2019.01.005
M3 - Article
AN - SCOPUS:85061701288
SN - 2210-6502
VL - 46
SP - 52
EP - 68
JO - Swarm and Evolutionary Computation
JF - Swarm and Evolutionary Computation
ER -