zerowaste_robot

Notes

the project was completed with compas_fab 0.25.0, from the ziv261_cfab250 environemt

On September 8th, 2022 (after completion of project) the visualizer component was updated: compas_fab 0.27.0 was installed from pre-compiled (unreleased) version, linked to rhino

Make sure to unlink and relink old version if needed in future...or just upgrade to official release now

Excerpt from slack convo:

• "the easiest option is to create an environment as usual, install compas_fab as usual from conda, and then clone the repo and run pip install -e . on the root folder of the repo which will replace the conda install with the source install after that, run python -m compas_rhino.install (as usual)"
• "let me send you a pre-compiled compas_fab from source + components" (this is so components get updated in GH)
• "So, here's a pre-release 0.27 compas_fab"
• "download it and then from the command line, with the conda environment conda_build_update activated, and then install with pip install compas_fab-0.27.0.tar.gz"

This was to fix an issue with ACM not showing up for robot2, wasn't fixed, but this was patched in a live session with Gonzalo

Issue was made on github about this:

compas-dev/compas_fab#372 (comment)

11/12/2022: Upgrade to Zivid SDK V2.8.1, the zivid2 no longer works through USB. Works fine through ethernet, but then have to change the network settings from the robot control setup. Using ziv281_cfab250 environment now

Folder Structure

Folder Name	Explanation
3dm	has the 3d model and the grasshopper script for structural analysis and path planning. Has RH custom display settings for MAC and WIN, and embedded files. Also contains a script that can be used to generate configs file planes for the calibration process.
configs	Holds the pre-saved robot configurations to move to locations for various robot move tasks: calibration, robot move presets, camera captures for shed/members/demo.
data	holds the data output from the camera captures. Individaul PCD captures and the stitched together PCDs for ECL_demo, members_phase1, members_phase2, shed_original, shed_new. For the HD versions of the PCDs see the ZeroWaste folder on Google Drive.
data_path_plan	the path planned robot configs for the various steps in each of the four disassembly/reassembly phases.
input_settings	zivid camera capture settings as yml files.
src	source code that is run from scripts in main folder.
transformations	camera transformation matrices to transform PCD into same coordinate frame, 4 step process.

Code Explanation

This is the code to run the robots and capture the PCDs of the structure

f1_rob_move_to_presets.py

• Move the robots to preset locations
• The presets are saved for each robot as joint values in the .json files in configs\presets\RX

f2_rob_cam_calibration.py

• Perform the camera calibration when mounted on to robots
• The goal here is to generate the H2 transformation from tool0 -> cam based on how the camera is mounted in relation to the tool0
• Two main functions here
• aquisition_calibration
  • captures the PCDs of the calibration board for the locations specified in the configs\calibration\R{} folder
  • can be run in save_config mode or in normal mode
  • save_config = TRUE, does not take any camera captures, just outputs the .json configs of the robots. Do this only once when you are first setting up the calibration locations you want the robot to move to.
  • save_config = FALSE, execute PCD capture, and save to data\calibration\R{} folder. Make sure the R1 and R2 folders exists here, currently not.
  • The aquisitions happen without transforming the PCDs, need them in original location as that is whole point of the calibration, saved as a {}.zdf file.
• perform_calibration
  • execute the calibration process after aquisitions completed
  • load the PCDs saved in the data\calibration folder
  • Use the code in the Zivid SDK, zivid.calibration.calibrate_eye_in_hand()
  • Each capture must have the checkerboard visible
  • Outputs the R{}_H2_tool0_cam file as the result of calibration to transformations folder
  • Note: there exist some .bat files in the data\calibration folder, which basically do the same thing as the perform_calibration function

There are three rob_cam_XXX tasks, that are similar in that they just move the robot to a series of pre-recorded places and capture PCDs:

• f3a_rob_cam_zerowaste.py: move both robots around the shed structure and capture this geometry
• f3b_rob_cam_members.py: move R1 to points around where the removed members are stored
• f3c_rob_cam_DEMO.py: move both robots to points around the little demo structure

f3a_rob_cam_zerowaste.py

• The main file to run the zerowaste robot + camera capture routine
• There are 3 main functions here
• aquisition_zerowaste
  • similar to the other aquisition functions, captures the PCDs of the shed structure at the series of locations specified in the configs\stitch_shed\R{} folders
  • performs a transformation on the captured pointcloud (so they all end up in the same coordinate system). R2 and R4 are the fixed transformations saved in transformations folder. R3 changes as it is the current frame of the robot represented as a 4x4 matrix in the pos{}.yaml file saved at the position of the robot at each capture.
  • saves the individual separate PCDs as {}_trns.ply file to the R1 or R2 folders in the data\stitch_shed folder.
• stitch_zerowaste
  • stitch together the individual PCD captures into a single file
  • Perform downsampling and other processing of the pointclouds, all set by the values of the variables. Some of these reduction functions are commented off.
  • can stitch together all the captures for a signle robot, if stitch = TRUE
  • can stitch together all the captures for both robots, if stitch_full = TRUE
  • For both can comment on/off either the High-Def or the Low-Def version of the combined file
  • Save as _R{}_pcd_stitched.pts for single robot, or pcd_full.pts for both robots combined in the data\stitch_shed folder. Can read these files into Rhino and manually clean and output at .ply