GPU processors for space surveillance and airborne radars. Towards highly efficient ARM-GPU architecture for Embedded radar systems

International audienceThe increasing complexity of radar systems and associated signal processing involves a significant increase of the computational workload. The challenge for ground-based radars is to reduce the cost of computing infrastructure whereas for airborne radars, the challenge is to implement intensive signal processing, such as STAP and SAR processing, while reducing hardware size and electrical consumption. In a concrete example of a space surveillance radar, we analyze all the signal processing modules running on CPU and on GPU, then we compare to the current computing system. For embedded applications, we present the advantages of a STAP algorithm running on GPU compared to a traditional general-purpose processor. We also show that using an original ARM processor-GPU configuration leads to an even higher efficiency than the classical x86-GPU architecture