AAAR 34th Annual Conference
October 12 - October 16, 2015
Hyatt Regency
Minneapolis, Minnesota, USA
Abstract View
SenseNet: An Outdoor Monitoring System for Biothreats
Ray Pierson, WILLIAM HARRIS, Cody Niese, Egbert Tse, Dave Wasson, Jonathan Thornburg, Quentin Malloy, Prakash Doraiswamy, Robert Serino, Northrop Grumman Inc.
Abstract Number: 630 Working Group: Bioaerosols
Abstract A scalable environmental monitoring system called SenseNet was designed to detect the presence of airborne biological hazards in outdoor urban environments by means of a distributed sensor network. SenseNet is a multi-tiered and multi-component system which combines mature, low-cost bioaerosol sensors with existing environmental, security and human activity sensors and data streams. The information gathered from the distributed sensor network is processed using innovative data fusion methods to provide greater situational awareness and rapid, high-confidence detection of outdoor releases of hazardous bioaerosols, enabling an informed and more effective response.
SenseNet is sensor agnostic and uses open-architecture network standards which allow the system to be configurable, adaptable and sustainable. Extensive modeling based on historical environmental data was used to identify the optimum architecture and sensor density. Modeling results included time-to-detect, probability of detection, and contamination mapping.
Modeled release scenarios from a group of six representative days for a “target city” using available historical data sets were “injected” with a modeled biological burst release. These scenarios were used as a training set for the system performance model. The plume PM mass concentration blended into ambient background within 600 meters (15 min) of release, meteorology dependent. The modeled ACPLA concentrations were 10$^4 when PM reaches background. As expected, the plume shape varied on meteorological conditions.
The system performance model was developed around a simulator which runs the learning algorithm for a given sensor configuration along with simulated inputs. The system performance model evaluated the diversity and density of the sensors, usefulness of the data, and data stream cost. The result was an optimum system design with a 95% probability of detection and suggested, low-cost collaborative environmental sensors reduced the need for the number of biological sensors.