In this chapter we have explored the different tools, algorithms and interfaces which can be used to work with point clouds in ROS. The reader might have noticed that we have tried to link the examples together to provide more insight into how these kinds of nodes might be used in a reusable manner. In any case, given the computational price of point cloud processing, any kind of architectural design will be inextricably linked to the computational capabilities of the system at hand.

The data flow of our examples should start with all of the data producers, which are the pcl_create and the pcl_read. It should continue to the data filters which are the pcl_filter and the pcl_downsampling. After the filtering is performed, more complex information can be extracted through the pcl_planar_segmentation, pcl_partitioning and pcl_matching. Finally, the data can be written to disk through the pcl_write or visualized through the pcl_visualize.

The main objective of this particular chapter was to provide clear and concise examples of how to integrate the basic capabilities of the PCL library with ROS, something which can be limited to messages and conversion functions. In order to accomplish this goal, we have taken the liberty of also explaining the basic techniques and common algorithms used to perform data processing on point clouds as we are aware of the growing importance of this kind of knowledge.