This chapter presents a study on robot perception. It begins by establishing the research motivation: enabling large-scale human detection and tracking in public (non-domestic) environments using embodied sensors and onboard computing. Subsequently, it introduces contemporary 3D lidar technology as an embodied sensor, covering its fundamental operating principles and relevant applications in mobile robotics. The chapter then details the “adaptive clustering” method developed by the author, highlighting its advantages and limitations through a performance comparison with other established methods. Following this, it describes several hand-crafted features extracted from point clouds, proven effective for human model training. Finally, it presents a multi-target tracker optimized for point cloud data.