Boundary Tracking and Rapid Mapping of A Thermal Plume Using an Autonomous Vehicle

Adaptive sampling algorithms for rapidly mapping the outflow plume from a nuclear power plant are presented. Plume detection is based on the notion of a plume indicator function that, loosely speaking, indicates the likelihood that an environmental sample is from the plume rather than from the ambient environment. For the case of a power plant outflow, we show that a plume indicator function composed of several measured parameters, such as temperature, salinity, dissolved oxygen, and flow magnitude, greatly enhance the robustness of our adaptive sampling algorithms. Successful field trials using a small autonomous surface vehicle are reported