Mutual Interference Avoidance of FMCW Automotive Radars Using AI

Recent years witnessed a rapid growth of FMCW radar applications because of their desirable characteristics such as small size, high range resolution, low peak power requirements, and cost effectiveness. In automotive industry the use of FMCW radar has become so prominent in Advanced Driver Assistance Systems (ADAS) as they allow robust detection in all-weather condition compared to cameras or LIDARS. Many radars are deployed on one car for different purposes such as Front Collision Warning , Rear Collision Avoidance, and Blind Spot Detection . Although there are large number of automotive radars are working simultaneously, there are limited spectrum resources allocated for this application at 24 GHz, 77 GHz, and 79 GHz bands. This leads to mutual interference between theses radars which might cause different effects such as noise floor increase and false target detections. A very important technique for interference mitigation called sense before transmission or Listen Before Talk (LBT) is used to sense the channel before transmission to avoid working in an occupied channel to prevent interference from other active radars. This technique requires blind detection of arbitrary parameters FMCW signals since there are no standardization of the modulation scheme or parameters used by automotive radars. In this paper, a robust blind algorithm, that detect arbitrary FMCW signal parameters under propagation effects, is introduced to address the aforementioned issues. The algorithm is based on a simple eight layers Convolutional Neural Network. This CNN is trained on FMCW detection task using WBR-DS-1 dataset.