The estimate is based on calculating the number of SLOC (Source Lines of Code) and it uses as a measure of the effort invested into code creation. The code is classified into one of five categories: the three “pure” and the two mixed intermediates. This classification is approximate because the authors have intimate knowledge of only one of the listed projects. It could be claimed that all projects contain libraries or other internal modules which are RSF independent. The fact that these libraries are not placed in DA category demonstrates an important point. To an outsider it is not obvious if such independent libraries exist and more important there is no indication that they will remain RSF independent in the future.
There is a notable exception within the reviewed RSS projects, the Kinematics Dynamic and Bayesian Filtering libraries distributed by the Orocos project and classified as pure DA code. The libraries which distributed by the OpenSLAM project the figure contain another block of DA code. These are shown as an example of robotic software which is not part of any specific RSS project.
The six RSS projects reviewed represent a significant portion of the global output of open source robotic software since the last 7 years. Put together they contain around 500 kSLOC. According to the classification, 67% of the total contains code which could be RSF independent and only 4% actually is.
Now it takes a step back from individual projects and tries to visualize how they all fit together. The marketplace of reusable robotics software as we see it today, While it has a layered horizontally structure at the level of operating system or OS and communication middleware, it is partitioned vertically at the level of robotic software. This partitioned structure reflects the common concern about the lack of inter-operation between different RSS projects.
No comments:
Post a Comment