wiki:Carologistics/Protocol/2018-09-25

Version 5 (modified by tviehmann, 4 weeks ago) (diff)

--

Carologistics Team Meeting September 25, 2018

Agenda

  1. Retrospective: Die Zukunft ist Heute in Austria
    • good organization, good field
    • footage available, including drone footage
    • moderation with 2 people on field (clearly visible for viewers)
    • other teams
      • ER-Force got C0 production running
      • Grips stabilized their performance
    • Carologistics
      • offset problems when positioning
      • gripping failed once and flipped the product away
      • gripping worked most of the times fine
      • relabeled cap station in refbox (cs1 to cs2), applying the wrong point cloud model to machines gives bad results
      • cap station problems like usual (wrong machine states)
      • Router crashed once
      • bug of exploration phase (getting stuck sometimes) got fixed
      • problem with one laser
      • real sense segfaults
      • goto got stuck when destination lays inside of an obstacle (fix this on skill level)
      • network simulation works (in simulation and real world)
    • event will be done again in the future (probably every other year)
    • Work for the future
      • robot memory
      • make the system more stable (also hardware wise)
      • speed up the vision
      • laser calibration
  2. Gripper Updates + Vision
    • Speed is an issue (currently >40 seconds for putting or getting a product)
      • ease up the alignment steps
        • reactive approach to prevent unnecessary alignments?
        • align gripper and robot at the same time?
    • Gripper
      • New design: 2 static fingers, spring to push the product against them
      • a lot of tolerance in depth
      • New Iteration: avoid gripping from above (not solved yet)
      • point cloud models at AGVR not working properly
        • Suggestion of Nick: record point clouds for debugging
      • Redsign the Arduino code
      • Speed up gripper motors in the future
      • use a planning approach for gripper movements? . mount real sense to the base of the gripper?
    • Vision
      • maybe use evolutionary learning approach
      • too slow currently
        • pipeline old vision, use as initial estimate for the new vision
        • enable real sense earlier
  3. Hardware Replacements
    • maybe new lasers + mounts, Nick tries to work some magic
    • maybe free 3D printing in the future (Daniel works on that)
    • Laptop of R-1 is causing problems (probably SSD)
    • have spare SSDs for every laptop
    • negotiate with Magazino for more sponsoring money/directly give hardware
    • Replace attachment of Tag Vision
    • Replacement for the wifi setup?
  4. Hackathon
  5. Playing Field Update
  6. Code Cleanup + Merging
  7. GitHub Migration