Non‐uniform virtual array position optimization for MIMO radar and neural network‐based radar imaging enhancement

Abstract This paper presents a method for enhancing the angular resolution of multiple input multiple output (MIMO) radar systems by employing an extended non‐uniform virtual array. Although a larger aperture achieved through non‐uniform antenna placement can enhance the resolution, it also introduces sidelobe artifacts that degrade the image quality. To construct a non‐uniform array that minimizes the sidelobe effects, the MIMO antenna placement is optimized using a genetic algorithm. Additionally, a convolutional neural network is employed to further suppress residual sidelobe effects that remain after optimization. Notably, to reduce the time and cost associated with collecting large‐scale training data, a simulation‐based radar data generation method is introduced. Both simulation and real‐world measurements using the TI AWR cascade radar module were conducted to validate the proposed method.