search

au-rmr / aurmr_tahoma

ROS workspace for our UR16e workcell
1 stars 1 forks source link

December 2023 evaluation updates #29

Open nickswalker opened 12 months ago

nickswalker commented 12 months ago

This bundles manipulation changes (manipulability scoring), state machine cleanup, and logging updates.

I have been able to run picks using this branch across both the realtime and perception machines, so this PR also ends those branches.

We experienced issues where the remote retry would not be invoked when running on Friday, and this branch does nothing that would fix that, so additional testing with a full stow/pick run would be helpful to uncover whatever is going on there.

@JackLowry some of your work in progress changes are here as unused states. Please feel free to include them in the state machine if they are ready to go

mmurray commented 12 months ago

We experienced issues where the remote retry would not be invoked when running on Friday, and this branch does nothing that would fix that, so additional testing with a full stow/pick run would be helpful to uncover whatever is going on there.

The issue where remote retry was not invoked was due to the dependency on this calibration file: https://github.com/au-rmr/aurmr_tahoma/blob/main/aurmr_tasks/src/aurmr_tasks/common/hri.py#L44

Soofiyan has a script that generates a new version of this file when we swap pods. However, he generates it on the perception machine while the state machine code runs on the real-time machine. So after re-generating the file on the perception machine it needs to be rsync'd to the real-time machine (that's how we fixed it on Friday). Or we could use a network drive to store and access the file.

nickswalker commented 11 months ago

@Soofiyan for future cases, this kind of calibration data (structured, order 100s values, potentially needs to be accessed by many parts of the system) would often be stored in yaml and loaded into the parameter server: http://wiki.ros.org/Parameter%20Server