Title: Multi-sensor Spatial Association using Joint Range-Doppler Features

Abstract: We investigate the problem of localizing multiple targets using a single set of measurements from a network of radar sensors. Such "single snapshot imaging" provides timely situational awareness, but can utilize neither platform motion, as in synthetic aperture radar, nor track targets across time, as in Kalman filtering and its variants. Associating measurements with targets becomes a fundamental bottleneck in this setting. In this paper, we present a computationally efficient method to extract 2D position and velocity of multiple targets using a linear array of FMCW radar sensors by identifying and exploiting inherent geometric features to drastically reduce the complexity of spatial association. The proposed framework is robust to detection anomalies, and achieves order of magnitude lower complexity compared to conventional methods. While our approach is compatible with conventional FFT-based range-Doppler processing, we show that more sophisticated techniques for range-Doppler estimation lead to reduced data association complexity as well as higher accuracy estimates of target positions and velocities.