Sparse sensor arrays for active sensing - Array configurations and signal processing

Multisensor systems are a key enabling technology in, e.g., radar, sonar, medical ultrasound, and wireless communications. Using multiple sensors provides spatial selectivity, improves the signal-to-noise ratio, and enables rejecting unwanted interference. Conventional multisensor systems employ a simple array of uniformly spaced sensors with a linear or rectangular geometry. However, a uniform array spanning a large electrical aperture may become prohibitively expensive, as many sensors and costly RF-IF front ends are needed. In contrast, sparse sensor arrays require drastically fewer resources to achieve comparable performance in terms of spatial resolution and the number of identifiable scatterers or sources. This is facilitated by the co-array: a virtual array structure consisting of the pairwise differences or sums of physical sensor positions. Most recent works on co-array-based sparse array design focus exclusively on passive sensing. Active sensing, where sensors transmit signals and observe their backscattered component, have been investigated less, despite their importance in ubiquitous applications such as radar and ultrasound imaging. The sum co-array naturally arises from the active sensing signal model, whereas the difference co-array is often more relevant in passive sensing. This dissertation proposes novel sparse array designs and signal processing methods for active sensing and imaging. We introduce linear and planar sparse array configurations that achieve a large contiguous sum co-array for diverse aspect ratios using significantly fewer sensors than conventional arrays. These low-cost arrays resolve vastly more scatterers than sensors in both azimuth and elevation, and synthesize beampatterns that are normally achieved by uniform arrays only. Several of the proposed configurations are symmetric, which implies that their sum and difference co-arrays are equivalent, and that they are suitable for both active and passive sensing. We also develop methods for coherent linear imaging, where image quality is improved by summing multiple component images, possibly corresponding to separate transmissions and receptions. We formulate a new optimization problem for achieving any feasible transmit-receive beampattern to a desired accuracy using as few component images as possible. We derive algorithms and closed-form expressions approximately solving this problem, and establish bounds on the number of component images of the optimal solution. We consider fully digital, hybrid, and fully analog beamforming architectures, as well as various waveform diversity cases, including phased array and orthogonal MIMO. Hybrid and analog beamforming further reduce the number of RF-IF front ends and related hardware costs, whereas waveform diversity governs the number of component images acquired per transmission. Numerical experiments verify the analytical results and characterize the trade-offs between the various system parameters. The contributions are of practical value in the design of sensor arrays for active sensing.