search

robotology / gym-ignition

Framework for developing OpenAI Gym robotics environments simulated with Ignition Gazebo
https://robotology.github.io/gym-ignition
GNU Lesser General Public License v3.0
220 stars 26 forks source link

Expose sensors #199

Open diegoferigo opened 4 years ago

diegoferigo commented 4 years ago

This issue is here as a placeholder to track the task of exposing the sensors supported by Ignition Gazebo to our ScenarI/O bindings.

Our external contributor @wolfgangschwab is currently trying to get the Lidar sensors working, facing quite a few problems.

Exposing sensors is not something we currently need with high urgency, but as soon as we find some bandwidth I would prioritize:

Something of our interest is also exposing and give the possibility to easily configure the noise parameters of all these sensors. In fact, this noise can be treated as a randomizable quantity #41.

diegoferigo commented 3 years ago

Relevant comment for the camera interface https://github.com/dic-iit/bipedal-locomotion-framework/issues/110#issuecomment-695196569 by @traversaro

diegoferigo commented 3 years ago

Relevant tutorial http://gazebosim.org/tutorials?tut=performance_metrics

diegoferigo commented 2 years ago

This issue is still open and there's not yet a clear plan to solve it. I still have to find a proper way to implement it.

Something that came in my mind is that this should be much easier if the ignition::sensors::<SensorClass> objects that are stored in the systems would be exposed to the ECM (e.g. see for the IMUs or the cameras).

This would require some support from upstream, but considering that in Ignition Garage there might be an effort to create official Python bindings, this could go in a compatible direction with upstream.

On top of them, some new ScenarIO sensors interfaces could be created and exposes so that we can implement sensors also for new backends when we will have more.