Hi all!
This is a great product and the effort from the developers and from the support team is faboulous.
HOWEVER
We, users, need stability, especially in the python API and in the documentation, and correctness of the simulation and of the expected behavior.
Every time you go out with a new version, the documentation pages simply disappear!
Every new version also old pieces of code simply stop to work out of nowhere, without a single deprecation notice or any way to get around the issue. In the latest code for example, viewport legacy disappeared auto_update_timeline does not work anymore, and the RTX lidar seems to be usable only through the replicator core module.
This means that either we users spend days trying to figure out how to update our code (since there is no document “this has been changed with this”, see “step_clock_on_physics”) or simply keep an old and less powerful version of the engine.
Please, try to be a bit more careful and helpful for us.
This has been a great product, but losing this much time to update a codebase because we need to figure out that the new system does not work as the old one without any single notice is not useful and will bring away customers!
I believe there is a bug where RTX Lidar may not add correctly from the UI.
For a workaround until a new version can be release, you can look at the standalone examples for rtx_lidar. there are ones for ros_bridge, ros2_bridge, and debug_draw. ./python.sh standalone_examples/api/omni.isaac.debug_draw/rtx_lidar.py
create hydra texture does work but ONLY within a SINGLE function (as soon as you exit that it does not work anymore). Indeed, in the test_rtx.py you use another way to create the render product
sensors.get_synthetic_data().activate_node_template("RtxLidar" + "ROSPublishPointCloud", 0, [render_product_path]) this does not even work (neither ROS, ROS1 or ROS2). Clean installation.
creating manually the graphs seem to not work either and there is no complete example (not even in the source files themselves).
Moreover:
cameras once are created and the simulation played, if you stop the simulation and re-play for whatever reason they completely stop publish the ROS topics (and nodes disappear from the graph apparently)
IMU are scaling twice the values (getting 98000 cm/s^2 for a 0.01 m simulation → 98 m/s^2)
timeline.set_auto_update(False) does not work (either before play or after playing the simulation). It advances time at every render so one needs to manually save the time render and go back
there is still no call that automatically fully render the scene (eg all the steps il pathtracing or accounting for FPS in raytracing, which is necessary when the render call is faster than the time necessary to compute the whole scene in complex scenarios)
manually publishing RTX lidars is cumbersome (difficult to understand buffer, the documentation says that they are captured at physics step but they are at render) and although one specify a rate they are not published at that rate but at the rate of .render() calls
manually publishing cameras is also cumbersome in some way with a very tricky .set(1) and .set(0) which, when compared to the old RosBridgeTickComponent is awfully complicated
syntheticdata getPose returns the wrong values for some components
3D bounding boxes are still buggy when one gets them from the synthetic data helper because there is no check of which objects are effectively in the rendered image (so at one line you can get an empty vector)