A MMW Radar Indoor Mapping Method Based on Transfer Learning

A millimeter-wave radar indoor mapping method based on Transfer Learning to generate dense detection data as Lidar’s output is proposed in this paper. This method uses the NN model with CycleCAN architecture to learn the Lidar-like map pieces, to enhance the mmw radar mapping performance. With the ideal of Transfer Learning, the model is trained using simulated data generated by CARLA and deployed into physical system to improve the mmw radar mapping performance. Simulation and practice measurement experiments are carried out to prove the coefficients of this method, and the quantitative analysis is conducted to evaluate the mapping quality.