Quantized Quickest Change Detection with Unknown Parameters Using Cloud MIMO Radar

Quantized quickest change detection (QCD) using cloud multiple-input multiple-output (MIMO) radar is considered for detecting an unknown change in a parameter vector from a normal to an abnormal state. Considering unknown postchange parameters, to reduce power consumption and hardware costs, observations collected by the cloud MIMO radar are quantized and transmitted to a fusion center (FC) sequentially. To deal with the unknown post-change parameters in the timevarying delays of MIMO radar returns, which result in timevarying statistical properties and high-dimensional and nonlinear challenges beyond traditional GLRT based QCD, we propose an extended quantized generalized likelihood ratio test (E-QuGLRT) based QCD method. To reduce computational complexity, we analyze drift property for the quantized observations for the first time and propose a drift-oriented QuGLRT (D-QuGLRT) algorithm. Theoretical asymptotic performance of the E-QuGLRT is analyzed, with closed-form expressions derived for the mean time to false alarm (MTFA) and average detection delay (ADD), along with an asymptotic performance bound for the worst-case average detection delay (WADD). The asymptotic performance for the D-QuGLRT algorithm is derived. We apply the E-QuGLRT method and the D-QuGLRT algorithm to abnormal respiratory rate monitoring using cloud MIMO radar, validating the theoretical analysis and demonstrating their potential for healthcare monitoring.