Real-time Implementation of Phased Array Radar Tracking Algorithm Based on Embedded GPU

The embedded Graphics Processing Unit (GPU) module, which includes both Central Processing Unit (CPU) and GPU processors, can be easily integrated into radar systems, offering high performance and flexibility. Phased array radar tracking algorithm involves numerous modules with poor parallelism between them, leading to significant data blocking issues during sequential execution. Traditional implementations using CPUs or DSPs are time-consuming and inefficient. To address this problem, this paper proposes a real-time implementation solution based on embedded GPU. The solution decouples logic-intensive and computation-intensive tasks in the algorithm and maps them separately to the CPU and GPU within the embedded GPU module. Then three-level hybrid parallel processing method is adopted, which is based on grid-stride-based intra-kernel thread parallelism, CUDA stream asynchronous parallelism and CPU thread parallelism in order to realize real-time processing of radar echo data. Verification experiment using real measured data indicates that the proposed real-time implementation solution achieves significant acceleration with no loss of accuracy on the Jetson AGX Xavier embedded GPU platform, markedly improving tracking efficiency.