E2DTF: An End-to-End Detection and Tracking Framework for Multiple Micro-UAVs With FMCW-MIMO Radar

Due to the weak radar echoes and strong background clutters in low-altitude airspace, the detection and tracking of multiple micro-unmanned aerial vehicles (UAVs) have posed formidable challenges in the radar surveillance field. Consequently, this article proposes an end-to-end detection and tracking framework (
$\text{E}\boldsymbol{ {^{2}}}$
DTF) for multiple micro-UAVs by utilizing the frequency-modulated continuous-wave-multiple-input–multiple-output (FMCW-MIMO) radar. To address the low signal-to-noise ratio (SNR) problem,
$\text{E}\boldsymbol{ {^{2}}}$
DTF presents a frame-range-Doppler-azimuth information fusion filter to integrate the target energy by exploiting the spatiotemporal dependence of positions within a sequence of unthresholded frames. In addition, considering that a target may enter/leave the radar field-of-view (FOV),
$\text{E}\boldsymbol{ {^{2}}}$
DTF introduces a target model state, updated by an extended Markov state transition matrix sequentially, to realize an unknown, time-varying number of micro-UAVs tracking. Another nice feature of
$\text{E}\boldsymbol{ {^{2}}}$
DTF is that it avoids the complex data association procedure thanks to removing the threshold-decision operation. Finally, both numerical simulations and experiments with five different scenarios, i.e., horizontal line, cross-trajectory, circular loop, rainy condition, and 3-D trajectory tracking, are presented to verify the effectiveness of the proposed method. The results show that
$\text{E}\boldsymbol{ {^{2}}}$
DTF can obtain superior detection and tracking performance for multiple micro-UAVs in contrast to the state-of-the-art methods considering detection and tracking processes independently, especially under low SNR conditions.