Prof. Francesco Mondada, Mobile robots architecture & design, Laboratoire de Systèmes Robotiques (LSRO), Ecole Polytechnique Fédérale de Lausanne (EPFL),
[caption id="attachment_896" align="alignright" width="178"] CC BY-SA 3.0 - Alain Herzog
In this extended guest post, Prof. Francesco Mondada describes a dilemma facing many researchers who use proprietary CAD software in their work but want to make their research results open: although education institutions, including universities, can get CAD software at much lower prices than commercial companies, such software comes with express restrictions on the dissemination of designs made with such software to industry. Although this rule in theory protects software vendors from misuse of educational licenses, it also means that CAD-assisted designs created as a result of publicly-funded research often cannot be made shared openly for access and re-use. After describing this situation, the author then presents a survey of the attitudes of software suppliers towards publication of designs produced using their software, possible solutions to the dilemma, and his recommendations for future steps.
Open Science isn't just about Open Access to publications amongst academics, but also about opening access to research data and designs in both academia and industry. Similar trends towards sharing can be observed in the maker community or in the electronics market, where open design is boosted by low-cost electronics, 3D printers, and laser cutters that everybody can afford and use to reproduce anything from simple objects to complex installations. It seems natural, then, that researchers should want to publish their mechatronic designs, such as experimental setups, as part of their open source research data. Considering that we are speaking about the researchers’ own intellectual property (IP), this should be just a matter of putting the files on a web server with the right distribution license, but, unfortunately, it is not that simple. In many cases, the dissemination of computer-aided design (CAD) files used for mechanics or electronics is limited by the license of the CAD software used to create them. It may be hard to believe, but in the case of hardware design, researchers cannot freely disseminate their own IP.
Their results are hence held hostage by the polices of software suppliers, and, in most cases, the only solution to secure free dissemination of the research results is to buy a commercial license which is much more expensive (sometimes by as much as 800 times) than the typical educational license universities are offered.
After two years of research on fish-robot interaction in the EC-funded ASSISIbf
project , our research group planned to publish, as open source hardware
, the full experimentation setup, including the mechanical and electronic design of the robot interacting with animals. The goal of this publication was twofold: to foster scientific progress by making these installations available to other researchers and to make this engineering work available to industry experts, whose taxes partly funded our research. This is also the core motivation of the European Commission’s Open Research Data Pilot.
On the OpenAIRE website, we can find the following definition: “Open data is data that is free to access, reuse, repurpose, and redistribute.” The mechanical design of an installation can be modified and reused only if given the original design files, where all the construction rules are set. For an electronic design, the situation is similar: to modify it, one needs to have the schematics and the generation mechanisms of the printed circuit board (PCB), which is the base plate on which electronic components are mounted, and, hence, a central piece of any modern digital hardware. Therefore, making these designs “open” means publishing the original source files that represent the design rules, as illustrated in Figure 1. Publishing illustrations, animations of the mechanics, or a production file in STEP (mostly triangle-based 3D) or GERBER (2D vector graphics) format does not allow reusing or repurposing the design.
[caption id="attachment_883" align="alignnone" width="758"] Fig. 1. Schematic view of the design process of an experimentation setup within the research process. Production files, illustrations, or even simulation files do not include design rules and cannot be used to modify and repurpose the design. The design information is only in the source design files, in the form of parametric 3D models. In most cases, these files are in a proprietary format, specific to the software used for the design, and they subject to its license.
In our case, the mechanical design of the ASSISIbf experimentation setup was made using CREO®
from Parametric Technology Corporation (PTC), a powerful mechanical CAD software. The product license is not explicit concerning the publication of files, but when asked about this, PTC clearly stated that transmitting files generated with an educational license to a company would violate the license, whether or not an open source channel is used. To enforce this, the educational version of CREO® even generates files that cannot be read by the commercial version sold to companies.
Format conversion software exists, but its use is restricted to users who have both commercial and educational licenses. Therefore, it is technically impossible for universities that own only the educational CREO® version to disseminate their designs in a way that allows companies to read them. This is clearly against the EC’s goals in its Open Data policies. As a result of this survey, PTC changed its policy, as described below.
The Basic Rules and the Global Trap
Behind this legal situation, there are a few simple rules that each seem reasonable on their own, but, when combined, result in a situation that seriously affects the free use of research results. The first commonly accepted rule is that education institutions, including universities, can get CAD software at lower prices than commercial companies, as they train future engineers to use these CAD tools and this is probably the best advertisement for the software. A similar rule is that commercial use of CAD software requires a more expensive license, which also makes sense. Combining these two rules, we get another general rule: designs made with educational software should never be transferred to commercial entities. Here we start getting closer to the essence of the problem.
This third global rule, which seems to protect the software companies from the misuse of educational licenses, implicitly forbids the dissemination of research results in industry. For instance, the license of Europractice
, an institution distributing electronics CAD software to universities, clearly states that “For the avoidance of doubt, none of the IP created using Europractice licenses may be licensed, sold, assigned or otherwise transferred to any third party, but shall remain vested at all times in your institute.” In the specific case of Europractice, the door is left open by the following statement in the license: “However, should a commercial situation arise as a result of work initially begun as a bona-fide non-commercial project, then Europractice will help find mechanisms to enable technology transfer.” Other software suppliers do not leave any such doors open and simply forbid any transfer, arguing that universities developing products for companies should purchase a commercial license. This is reasonable, but, in the specific case of open source publication, it requires that researchers who simply want to publish their research results must purchase a commercial license.
[caption id="attachment_897" align="alignleft" width="179"]
CC BY-SA 4.0 - Lorenzo Natale
One of the largest open source hardware projects in robotics, the iCub
humanoid robot, got trapped in exactly this situation. To be able to publish their robot designs, the academic developers were forced by the CAD supplier to purchase a commercial license, which was 500 times more expensive than the educational/academic one. This was the company’s requirement to allow free publication, with no additional services provided. Moreover, since the file format is proprietary, any company that wants to read the files needs to purchase a commercial license for the same software. This doubles the profit for the software producer, taking advantage of the dissemination of research results that should be accessed for free! The iCub researchers managed to find a way to purchase a commercial license and finally published their results, but what about all other research projects? Should researchers pay such a high ransom to disseminate their results?
When asked about this issue, the international property rights (IPR) HelpDesk of the European Commission and the legal department of the École Poltechnique Fédérale de Lausanne, where the ASSISIbf design was created, confirmed that this is a serious restriction of the use of intellectual property. Both legal offices came to the same conclusion: researchers should avoid using software having such restrictive licenses.
Indeed, the immediate solution is simple: universities should not use software with a license that implicitly or explicitly forbids the publication of their results. Table 1 summarizes a survey we performed among the main CAD suppliers over the last few months, and it showcases their very different stances toward publication. Among the suppliers of mechanical and electronics CAD software, several have educational licenses that allow the free publication of results, and these have explicitly confirmed to us that commercial use of the results by somebody downloading the files would not violate the educational license. Other suppliers explicitly request that the publication of the files be associated with a known open license, such as the Creative Commons (CC) or the license of the Open Source Hardware Association (OSHWA
), which is a very reasonable condition. Finally, some simply forbid publication to organizations that have not purchased an expensive commercial license.
[caption id="attachment_887" align="alignnone" width="799"] Table 1: The surveyed software suppliers, their CAD software, and the policies they apply to the publication of source files. The suppliers highlighted in green have a very open policy, allowing a commercial organization to download files published by researchers working with educational licenses. Thoses in yellow have licenses not allowing such publications, but explicitly mention exceptions for those who ask for them. Those in red simply do not allow or did not confirm the possibility of publication of files generated with an educational license if made available to commercial entities
The large companies like Siemens and Dassault simply delayed for months their answer with various reasons. Only in the case of PTC, after a long discussion involving the whole corporate hierarchy, including the CEO, the company understood the catastrophic impact of its existing licenses on the publication of research results. Unable to change their licenses in the few weeks before this publication, they immediately decided to grant automatic approval to all requests for the publication of explicit open source (under a CC or OSHWA license) and European research projects designed with educational licenses. This applies to both educational and commercial formats, and PTC agreed to provide conversion software if needed. This shows that an evolution of the existing situation is possible, even in a short time.
The Next Steps
Having the right CAD software license solves part of the issue, but not the whole problem. Indeed, these CAD files, even if public, are still in proprietary formats. Some suppliers provide free readers, but the real use of the design and its modification still requires purchasing the original CAD software. This seems unavoidable for the time being. Unlike the use of office applications, using free open source software for mechanics and electronics designs still seems unfeasible as open source alternatives, such as FreeCAD
for mechanics design, are still under development. This is a common discussion in open hardware communities like Arduino
. The Arduino open hardware design is made and distributed using the Eagle®
proprietary software, and many members of the community have already advocated for the use of open source CAD software. The core designers still prefer using Eagle®, as it is far more efficient than the open source alternatives.
Ultimately, a more lasting solution would be to have a common open file format for the encoding of mechanical or electronics designs. This poses a daunting technical challenge, as the representation of the geometrical models used for CAD software are tight to the underlying mathematical representation of the geometry engine, which varies from software to software. The International Organization for Standardization (ISO) is working to improve the STEP format (ISO 10303), for instance, by including kinematics or extending the scope to electronics.
However, this updated format does not solve the problem of maintaining the edition tree of the geometry; therefore, it is not a solution to the core problem. Let us hope that, in the future, CAD suppliers, standardization organizations, open hardware developers, and members of the research community will team up to create open formats that will finally set this valuable research data free.