Clean up new urdf

[fiorisi]

This improves handling and generation of the models (URDF, SDF, DH parameters and meshes) that we use for R1.

The main problems with the previous repository were: 1) the urdf, the sdf and dh models must represent the same robot. Sometimes changes in the sdf model were not propagated in the urdf or DH parameters (or vice versa); 2) changes in the sdf and urdf are lost during the regeneration of the models. The resource files (the one used to generate the models) had to be updated by manually searching the changes in the repository history; 3) the models need to be tested to check their consistency with the CAD model. This check was done manually (e.g. DH test ).

The new repository has the following differences:

• resources: this folder contains all the resources necessary to generate the models. The changes to the models now take place by modifying the resources files;
• dh_parameters: this folder contains the parameters of the DH chains;
• generate_models.py: this script deletes the old models and generates the new one using the files inside of the resources directory.

Thanks to this approach it is possible to solve the problems 1) and 2). In order to solve also the third problem we can write some simple tests to run on a continuous integration service. If we want to use Travis for free we need to make this repository public.

Other changes inside of this pull request:

• deleted the old cer models
• updated the model with the new frames and plugins
• updated readme.md

@drdanz @randaz81 @pattacini @barbalberto @ale-git @aerydna

[pattacini]

@fiorisi I think at this stage there's nothing preventing us from making the repo public by now.

@lornat75 @miggia what do you think?

[miggia]

we kind of started a discussion on this with @fiorisi but coul'dn't finish it because lack of time. I don't like being encouraged to choose to make a repository public by github or its surrounding ecosystem (read Travis) by their pricing policy. I think the simulation model is something that might contain information that can be considered competitive advantage (link masses, joint torques, actuator efforts), and I'm not very happy with its disclosure. As for the testing we could consider running it prior to pushing changes without Travis. I don't see the clear benefits of continuous integration testing in this case. On this I'd really like @lornat75 opinion; eventually even Giorgio's one. Maybe I'm just being paranoid, but given a choice I prefer the conservative option.

[pattacini]

@miggia actually, disclosing cer-sim shouldn't be related to using Travis or other CI tools. It's be rather a matter of how we see our future development, that is if we think we're going to open source the software, then these models will need to be definitely public in order to let users run Gazebo to test the code and release their contributions.

[miggia]

Ok, let's drop the discussion on Travis.

Since the robot is so far only being used by us I see no point in making the repo public.

Before making the repo public I'd like to us to revise with @lornat75 and possibly @giorgiometta the disclosure policies for the robot. My impression is that we never really took time to consider in detail these issues; I think we should conceive a general "strategy" and stick to it.

[pattacini]

👍 for the discussion.

However, this issue is not the good place to do that. Let's go on reviewing @fiorisi's PR and let the other guys comment and eventually accept it.

[randaz81]

Regarding the above discussion on the repo, I have something important to tell you privately at my desk. @pattacini @miggia @fiorisi.