High-performance image reconstruction algorithm in CUDA C++ for ultra wideband multi-channel MIMO radar systems

The exact measurement of process-relevant parameters and product properties are prerequisites for efficient and sustainable production. In addition to accuracy, industrial applications place tough demands on the real-time capability and achievable measurement rates of the sensor technology. In the past, radar signal processing was mainly done with the use of highly specialised hardware to achieve the necessary performance. Computer systems are used to perform simulations and to test new algorithms before being implemented under high effort. The resulting sensor systems are rigid, and their enhancement is time and cost consuming. With increasingly powerful graphics processing units (GPU) and the possibility to use them for general purpose computing, a new approach is to outsource parts of the radar signal processing from the specialised hardware to commercially available computer systems. The main objective of this idea is to reduce the development time of new sensor systems, facilitate their modification and to increase the re-usability of produced code. This approach is tested with a new imaging radar algorithm, developed for a frequency modulated continuous wave (FMCW) radar system with a modular multiple input multiple output (MIMO) antenna array. The implementation of this algorithm is used to determine the boundaries of this new approach and involves a step-by-step optimisation process to improve the performance of the final result.