Re: Graphical programming support
Peter Soetens <[email protected]> Mon, 11 Dec 2006 15:25:26 +0100
| Newsgroups | gmane.science.robotics.orocos.user |
|---|---|
| Organization | FMTC |
| Message-ID | <[email protected]> |
On Saturday 09 December 2006 15:58, [email protected] wrote: > Dear Sirs > I am a member of a team managed by Prof. Gianantonio > Magnani of Politecnico di Milano. > > Our team works on control systems > for robotics and related fields. > > I am using OROCOS since release > 0.16 and I think that OROCOS is very powerful, > but beginners have to > do strong effort to acquire the knowledge necessary to use the > framework. > This prevents a wider diffusion of this powerful and useful > tool. > There are users that don't have a deep knowledge of C++ and > don't need to develop > new components, often they only need to > concentrate on assembling control blocks already available. C++ is what makes this framework powerful, but also what makes it difficult to the 'matlab' control engineer. Although we know out of experience that such kind of people can be teached to use it in a reasonable amount of time. We're lacking an 'Orocos Basics in 20 minutes' manual, which _is_ possible, once it is installed. But you're right, it ain't point-and-click. > > Our > group is thinking about a friendly user interface, which should allow > this category of users to > build control systems from available blocks: > controller, sensors, kinematics chain and so on. > > This user interface > in our idea should expose component as graphical blocks that can be > dragged > on canvas and connected each other in a graphical way. > Users > should only specify relevant parameter for each block. > The interface > should generate the correct C++ code based on OROCOS framework from > the > raphical model. > The goal of this interface is to avoid the effort > of writing C++ code to user focused on control. > Our goal is to build a > tool that allows faster and easier development of control system and > to > assemble a library of components. If possible, you could extend an existing framework, which would give you a larger pool of testers. Herman mentioned 'Marie', although it is again a framework for mobile robots. Go carefully through the projects listed here: <http://www.orocos.org/node/29>, but we'll add yours to the front page if you start a new one :-) On the other hand, if you'll only create a 'code generator+frontend', starting from scratch won't be that bad. > > Another field of interest is > the definition of robotic systems, like kinematical chain, using a 3D > cad > like Blender. > We are thinking about the development of a second > tool that should generate the C++ code > for a complete robot, based on > OROCOS framework. > This tool should use a 3D modeler to define the > kinematical chain of the robot and a graphical interface to complete > the > model with information about sensors, controller and drivers etc. > > This information is really succinct but I hope can give an idea of > our goals > and we are very interested in your opinion about these > activities. I really think this can succeed. I know that all of these kind of projects have 'lost' a lot of time in getting the bugs out of the 'lower' layers. Designing _both_ a GUI and a robot controller is a _lot_ of work. Using a 'generator' which relies on a mature library is surely a way to have quick success. Peter -- Peter Soetens -- FMTC -- <http://www.fmtc.be>