Research Summaries

Back Sensor Fusion for Undersea Operations

Fiscal Year 2018
Division Research & Sponsored Programs
Department Naval Research Program
Investigator(s) Yoshida, Ruriko
Szechtman, Roberto
Sponsor NPS Naval Research Program (Navy)
Summary We propose to address certain important problems to perform probabilistic and statistical analysis in an information fusion setting. A typical operational scenario will have multiple sensors deployed collecting information, so that their observations need to be brought together in order to rapidly detect relevant events.
A sensor may operate alone or form part of a larger ensemble that jointly reports on the situation of interest, in which case a fusion center gathers data from each sensor, processes the aggregate data, and may provide feedback to the sensors. The communication between sensors and the fusion center is often subject to bandwidth constraints, meaning that sensors compress the incoming signal before transmission to the fusion center and vice versa. In the simplest scenario, the analyst wishes to ascertain which of two hypotheses, H0 and H1, is true. Given a prior on H0 and H1, the goal of the fusion center is to reach a decision that minimizes the expected total cost, where cost accrues from rejecting H0 or H1 when true and from the time it takes to reach a decision (given by the cost times the number of samples per sensor until a decision is taken).
Until recently most sensors had a human in the loop - for example, checking a video feed from a security camera or interpreting spoken language from intercepted communications. Recent advances (e.g., in image and speech processing) allow for improved autonomous sensing. In either case, the sensitivity and specificity of the sensor improves with experience and training - for instance, by having an expert identify a person of interest from video frames -- but this approach is costly and the improvement eventually plateaus. Related to this is the fact that, for sensors that are untrained or fielded in novel operational settings (e.g., when mounted on unmanned undersea vehicles patrolling a new area of interest), there is uncertainty about the data stream densities.
Keywords
Publications Publications, theses (not shown) and data repositories will be added to the portal record when information is available in FAIRS and brought back to the portal
Data Publications, theses (not shown) and data repositories will be added to the portal record when information is available in FAIRS and brought back to the portal