Adding robot model to Towr

popi-mkx3 / popi_project

Here is everything you need to know about POPI, our open-source quadruped robot. If you want to check the videos we will release about it, you can have a look at our YouTube channel.

https://www.youtube.com/channel/UCaCy-1MX6SoqdtjZH4uU_TA

BSD 3-Clause "New" or "Revised" License

109 stars 26 forks source link

Adding robot model to Towr #4

Open Mikipaw opened 3 years ago

Mikipaw commented 3 years ago

You said that you easily added your robot model in towr. Can you please explain how to do it with my robot unitree a1?

popi-mkx3 commented 3 years ago

Hello Mikipaw ! Thanks for reaching out.

I suggest you have a look here to see how different robot models are built.

You could also clone the repository, go to _catkin_ws/src/popi_project/popisoftware/towr and then use grep -rnwi . -e "popi"

This way you should have more insights on which files we added or modified to have our model working.

I don't remember stating anywhere it was an easy process though, and can't provide a step-by-step method on how to add your own model. You should also have a look at the doc written by Alexander W. Winkler for Towr.

If you have any more questions, please provide some details on what you've tried so far and what problem you encountered. A good way to do this would be to fork Alexander repo (or this repo) and commit your changes so that I can have an in-depth look at your work. It will prove really difficult to help you otherwise.

Hope this will be of some help for you.

Cheers, Lucas

Mikhalinho commented 3 years ago

Hello again!

Can you please give me an example of 3D map where popi creates path using towr?

And how do you send the map to popi?

popi-mkx3 commented 3 years ago

Hi ! Are you the same person ? Did you have any success following my previous message ?

I think you are mistaking about what we did with Towr. We actually didn't do any SLAM or path planning. We used Towr to create joint trajectories for a simple walking gait. Then we saved those trajectories and sent them to the real robot. You'll have more details on the whole process in this README if you want.

Hope this helps. Lucas

popi-mkx3 commented 3 years ago

If you want to have a look at the trajectories created by Towr, here is how we did it.

When running roslaunch towr_ros towr_ros_popi.launch, here is the rqt_graph

Basically here is what is going on :

xpp computes desired states given cost functions, max deviation of the foot from the body, etc. It gives you desired feet positions.
the urdf_visualizer_popi4 actually reads your URDF and this is where IK is applied to get joint angles. It allows you to see things in RViz. But this wasn't too convenient though, as it was in a xpp_msgs::RobotStateJoint type of message.
So we wrote a small node, joint_pose_translator to separate the joints positions and send them to the controllers.

Then, using rqt_bag, we could monitor these topics and save the trajectories for each joint.