loading the same controller 2 times?

[mitsui29]

Hi, thanks for your nice projects! I have a question for the dual_arm launch file:

there are 2 arms which will be used. this means that we have 2 robot_id, and so the same controller bimanual_cartesian_impedance_controller will be loaded 2 times, event though 2 control inputs can be computed in one bimanual_cartesian_impedance_controller.cpp file at the same time. As result of this 2 times loading, the 2 control inputs will be computed 2 times. Is that correct? I open this issue, because i have to save some computational sources. is that possible, move out this controller from the namespace(ns) to public scope? Thanks in advance!

[franzesegiovanni]

Hi!

I will give it a look the coming week. I just saw this message. I am not sure if that is possible.

Do you get some communication constraint violation?

Giovanni

[franzesegiovanni]

Hi! I don't think that the controller is loaded two times... Can you be more specific with the error you get and for the system you have? I also had some issue with Fr3... I used pandas up to now.... Are you also using fr3?

[mitsui29]

Hi Giovanni, Thanks for your replying! I think the issue I have stated may have led to a misunderstanding. As you said, the controller didn't loaded 2 times. It's correct. Cauz, at that time point, i thought maybe you can also run this project on the Gazebo. But i couldn't run it on the Gazebo. This is why i opened this issue. Actually, i just wanna verify my proposed control framework to control 2 Franka pandas for some tasks. But what i proposed control algorithm is not decentralized. That means i have to use one control file to control 2 pandas on the Gazebo. After searching some issues on the Github and having a look on the FCI Website again, i think there are no ways to verify my ideas on the Gazebo. The reason is that there are no supports for Franka_Combined_HWSim(Simulation version for franka_combined_hw). The errors are mainly related to interface resource conflicts and failure to find the controller. My system 20.04 noetic, gazbeo, franka panda. Maybe our issues are not same...

Zihao

[franzesegiovanni]

Hi, I think that could be a way of testing it on gazebo. I will check it out and come back to you.

Kind Regards,

Giovanni

[franzesegiovanni]

Hi,

I had no luck trying to figure this out. There is no clear way on how to simulate the combined hw. Let me know if you find a way.

Kind Regards!

Giovanni

[mitsui29]

Hi, Thank you for taking the time. I've spent about a couple of days trying to solve this problem before, and it doesn't work just from a gazebo file configuration perspective, I think the most direct way is to rewrite the simulated interface like franka_HwSim, but this part is hard-coded and involves a lot of work, I read a similar answer from the maintainer of this justagist/panda_simulator source, because he provided the gazebo simulator before the official franka, and one of the main tasks is to write the simulation interface. I didn't find a suitable answer by searching the resources on github. It seems to be a problem that still plagues some people at the moment. If you want to test this kind of centralized control on a simulation platform, mujuco seems to work (not 100%). Of course if I know how to solve it, I'll let you know asap.

Kind Regards!

Zihao

[franzesegiovanni]

Hi Zihao,

yes indeed, I had the same feeling. We must change the franka hw Sim to simulate two frankas with a centralized control. I usually do not rely much on the simulator, I prefer to test directly on the real robot. Good luck with it and I am looking forward to a solution to the problem.

Kind Regards,

Giovanni