Design of Reconfigurable Radar Signal Processor for Frequency-Modulated Continuous Wave Radar

Radar sensors are utilized in a variety of applications owing to their robust performance in harsh environments with minimal environmental constraints. Target detection is a crucial aspect of utilizing radar sensors. Radar signal processing for target detection entails fast Fourier transform (FFT) and constant false alarm rate (CFAR) operations to determine the distance and relative velocity of the target. However, the computational intensity of FFT and CFAR operations poses challenges for real-time operation. Despite the need for diverse radar specifications across different applications and operating environments, traditional radar signal processors designed for specific uses limit versatility. Therefore, this article proposes a reconfigurable radar signal processor (R-RSP) comprising FFT and CFAR intellectual property (IP) modules to dynamically accommodate different radar requirements. The FFT IP can support FFT operations on variable-length inputs, ranging from 64-point to 4096-point inputs. Similarly, the CFAR IP supports various algorithms, such as cell averaging (CA) CFAR, ordered statistics (OS) CFAR, smallest of OS (SOOS) CFAR, and greatest of OS (GOOS) CFAR, providing flexibility to select the optimal algorithm for different environments. To validate the R-RSP, we conducted detection experiments involving humans, drones, and cars utilizing a variety of radar sensors. The results demonstrate a minimum acceleration of approximately 16 times and a maximum acceleration of approximately 1130 times compared to the performance of the ARM Cortex-A53 microprocessor.