SSC-RS: Elevate LiDAR Semantic Scene Completion with Representation Separation and BEV Fusion