How would you approach 3D point cloud processing for autonomous perception in Autonomous Solutions systems?