Closed ZhiyeTang closed 2 years ago
That's a good pickup @TongChiYip ; thanks for that.
This is actually a bug in our updated controller code. The pose 'camera'
should exist in that object.
I'll make a fix and push out a new release later today.
Above commit should fix this. Let me know if updating BenchBot doesn't resolve the issue.
Thanks @btalb . Is a complete reinstallation needed? Or we just need to rebuild the Docker images related?
If you just re-run benchbot_install
it will go through all the steps, skipping whatever is already installed and reusing cached Docker layers.
Due to the firewall set by our government, a re-run of benchbot_install
might be annoying, which makes me installed benchbot
in a "semi-manual way" previously. If a simple re-building for related Docker images is all what I need, that would save my time.
Ah I see, no worries.
In that case, yes updating the Docker images should be all you need (mainly Git pulling the repo in /benchbot/benchbot_robot_controller/
directory, then committing the ).
You may also want to pull the latest API in the benchbot/api
directory outside of Docker as well.
So the updates are only related to Git repos benchbot_api and benchbot_robot_controller?
That's correct.
I've rebuilt the related backend
image and simulator
image. But when I try benchbot_run
, another error occurred showing that the BenchBot Robot Controller container exited unexpectedly due to file /benchbot/ros_ws/src/benchbot_robot_controller/nodes/benchbot_robot_controller
is not executable. This error haven't happened in previous versions. Could this be caused by the catkin compilation? Or just my mistakes? I'm not sure.
That I would believe be an error somewhere in the manual process you've taken @TongChiYip.
That file's existed for a long time, and not been changed in any recent updates. If you're re-building with Catkin manually, you should just have to source /benchbot/ros_ws/develop/setup.bash
, re-run catkin_make
, and then Docker commit the results.
Sorry I can't offer anything more concrete, supporting manual building is a bit outside of our scope.
Sorry for bothering you again. I was trying to build my first agent according to your tutorial, a
KeyError
occurred showing that there is no key called 'camera' in the dictionaryobservations['poses']
: I think it might be caused by some API mistakes, while there is actually no mention for a key 'camera' according to the README file in benchbot_api repository.