- Error in perception module:
"error processing request: conv2d(): argument 'input' (position 1) must be Tensor, not NoneType"
- Reason: To accelerate computing, the rendered images are cached. Sometimes the cache is not complete. You need to delete all the cache.
- Solution: Delete all files in
/root/ocrtoc_ws/src/ocrtoc_perception/src/ocrtoc_perception/pose/rendered_object_images
EXCEPTviews.yaml
.
- Error in moveit:
"PluginlibFactory: The plugin for class 'rviz_visual_tools/RvizVisualToolsGui' failed to load."
- Solution: It doesn't matter.
- Error in x server when debug is set to
True
while usingsuperglue
perception method.
- Reason: This problem is caused by pyrender. After rendering images, open3d and matplotlib visualization can not be used.
- Solution: Currently, you need to delete all the visualization after rendering.
- Why Sapien simulator stops for a short time each second.
- Reason: Sapien renders and publishes images at a frequency of 1 Hz. As a result, when the graphics engine is rendering, the physics engine waits until it ends. However, this doesn't affect the phisical process.
- Error in Sapien simulator:
"A controller named 'position_joint_trajectory_controller' was already loaded inside the controller manager"
- Solution: It doesn't matter, you can continue using the simulator.
- Can I read the object 6d poses from simulators?
- Answer: You may use the
/get_model_state
interface while debugging. An example is inocrtoc_task/scripts/get_6dpose_dump_simulator.py
. But you should NEVER do that in your solution. The scores will be cancelled if we find that you read the poses directly from the simulator.