Tutorials about python bindings

[zichunxx]

Tutorials about python bindings

[zichunxx]

After browsing the official demos, do Python bindings not provide the function to add mesh obstacles? Only ompl.app can?

[ChamzasKonstantinos]

Hi @STAY-Melody the ompl library (both C++ and python bindings) don't have an option to add geometric obstacles. This is always done implicitly by specifying a collision checking routine provided by the user. Check this link which explains the justification behind this choice and how OMPLapp/MoveIT does it. (which are extentions of OMPL)

[zichunxx]

Hi @ChamzasKonstantinos

Thanks for your reply.

As you mentioned, the only way for Python bindings to finish the motion planning task with obstacles is to define the real environment through state validity checking.

However, is this method easy for primitive-shaped obstacles and difficult for complex shapes?

I want to finish the motion planning task only with Python and without ROS and MOVEIT. Is it possible?

[ChamzasKonstantinos]

It is much easier to write you own function for simple primitive shapes. For complex meshes you will usually need a collision checking library like FCL.
If your robot is a rigid body (e.g. a car or a drone but not a manipulator) you can try creating mesh files that are compatible with OMPl.app (one for your robot and one for the environment) and do your planning through OMPL.app. You can see an example in python here. For manipulators unless you write your own kinematics and collision checking routitnes, you will need to use MoveIt or something equivalent.

[zichunxx]

Thank you so much. I will follow your introduction and go deeper.

[zichunxx]

BTW, are there some demos about using FCL to construct the validity checking function or designing a planning task with multi-DOF mechanisms?

[ChamzasKonstantinos]

This is the way OMPl.app uses it: link
You can also check other software that use OMPL under the hood here maybe one of them would suit your purposes better.

[zichunxx]

Thanks