Information Theoretic Design of Sparse Arrays and Adaptive Algorithms to Maximize Detection in Real Sonar Systems
Abstract
The Navy has a strategic interest in being able to detect, track, and localize extremely low SNR sources in the presence of elevated noise levels with various interferers using passive sonar systems. Sonar systems estimate the location of an object of interest by processing pressure observations received by a hydrophone array. For a given number of sensors, sparse sensor arrays provide more degrees of freedom than a full array. The enhanced degrees of freedom can be exploited in detecting and localizing a greater number of sources than sensors. Many different sparse arrays have been explored for decades, such as minimum redundant arrays and logarithmic arrays. The recent introduction of nested and coprime arrays has sparked a renewed interest in the sparse arrays category. The primary advantages of nested and coprime arrays are the availability of closed form expressions for sensor locations and the existence of clear mechanisms to disambiguate aliasing artifacts resulting from undersampling. For a given aperture, it is possible to design many different sparse array geometries. There are two primary challenges involved in the implementation of sparse arrays for real sonar systems. The first challenge is the absence of definitive criteria that indicate which sparse array geometries, or their combinations thereof, are best suited in extracting information from real sonar data. One objective of the proposed research is to overcome this challenge by using information theory measures, such as Kullback Leibler divergence and Shannon entropy, to characterize sparse arrays geometries and their combinations. The second challenge is that an overwhelming majority of the existing research on sparse arrays has been limited to non-adaptive algorithms, such as conventional beamforming, product processing, minimum processing, multiple signal classification, and compressive sensing. The signal and noise models for sparse arrays have been restricted to plane wave signals at design frequency, embedded in spatially and temporally white noise, as required by non-adaptive algorithms. Such simplistic models are not useful for the Navy because these models do not accurately reflect the highly variable and sophisticated environment in which real sonar systems operate. Adaptive array processing algorithms are required when realistic sonar signal and noise models are considered. These algorithms adapt the weight vectors to the measured data and provide better detection and localization of quiet sources by nulling out noise and interference. Another objective of the proposed research is to overcome the second challenge by designing adaptive algorithms for sparse arrays operating in a real sonar environment. The project will assess the performance of the adaptive algorithms using information theory bounds, such as Shannon capacity bounds and rate distortion bounds. The specific outcomes of this project are (1) identification of the sparse array geometry that provides the best detection of weak signals in a given scenario using information theory tools and (2) robust adaptive algorithms applicable to sparse arrays. As a result of this project, the Navy will know the precise strengths, payoffs, limitations, and weaknesses of these sparse arrays before implementing them into real sonar systems, thereby enhancing DoD s overall technical capabilities in alignment with the mission of DEPSCoR.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 08, 2020
- Source ID
- N000142012448
Entities
People
- Kaushallya Adhikari
Organizations
- Louisiana Tech University
- Office of Naval Research
- United States Navy