Optimization of Distributed Netted Sensor Fields with Application to Undersea Surveillance

Open Access
Katic, Megan Christine
Graduate Program:
Industrial Engineering
Master of Science
Document Type:
Master Thesis
Date of Defense:
Committee Members:
  • Dr Soundar Kumara, Thesis Advisor
  • Soundar Rajan Tirupatikumara, Thesis Advisor
  • Jeffrey J Weinschenk, Thesis Advisor
  • Distributed Netted Sensors
  • Undersea Surveillance
  • MATLAB Simulation
Distributed-Netted Systems (DNS) can be used in many different environments. In particular, several Undersea Distributed-Netted Sensors (UDNS) have been explored in software simulation and hardware prototypes. While specific approaches have been studied in great detail, the literature lacks a broad view that explores which sensing modalities and sensor field compositions yield the best performance. The literature fails to answer questions such as is mobility cost effective, how fast is fast enough, are fewer high performance sensors preferred over many less expensive lower performance sensors, etc. We introduce a novel geometric method that allows rapid analysis of competing approaches including heterogeneous mixtures of several different sensor types. Our approach will allow planners to select Pareto efficient solutions for undersea surveillance applications. Within our approach, the sensors are modeled as disks which represent their acquisition footprint. The sensor will be able to detect, classify, localize, and track (DCLT) any submerged contact that enters this region; we refer to the sensor‟s ability to DCLT as "acquiring‟ the contact. Submerged contacts can only enter along one side of the rectangular area, known as the border. The contacts are described in terms of their offset from one side of the region and the angle they create with the border (also referred to as the heading). Probability density functions are used to describe these offsets and headings, and their combination results in a probability mass function (PMF). A coverage function is developed to determine which of these offsets and headings are "covered‟ by a sensor located at a particular point in the area. The goal of the approach is to maximize the probability of acquisition, which is calculated as the Frobenius inner product of the Hadamard product of the PMF and the coverage function. The probability of acquisition of a DNS is plotted against the cost of the DNS and a Pareto analysis is performed to determine which DNS should be implemented.