wiki:Carologistics/Protocol/2019-02-07

Version 1 (modified by tviehmann, 12 days ago) (diff)

--

Recap on 2018

  • Wifi issues
    • freezing AP issues
    • solved by using laptops directly
    • probably stick to that in 2019
  • Slow assembly in picking and putting
  • Old Gripper
    • unreliable wp position inside gripper
    • bad positioning on the belt
  • Started too late
    • it is getting better
    • meet regularly
  • Agent
    • Needs testing in real world

Currrent Status

  • Agent
    • Main Task: Multiagent
      • Needs real world testing
      • Maintenance does not work (syncing worldmodel)
    • Mongodb
      • Switch from legacy to new driver
      • Not open source anymore -> will be removed from fedora, mongodb provides rpms and we wont have legal troubles
    • Multiagent Strategy should work overall
    • Work on low level feedback
    • Exploration is expected to work
      • remodel dirty hacks
    • Overall mainly syncing issues
    • Lock expiration will help to avoid getting stuck
  • Skills
    • In progress of merging and cleaning up skills from RoboCup2018
    • Adapt to new gripper, changed camera setup and other hardware changes
    • Morian is currently the only one working on skill
    • Need a lot of optimization all around
    • Start from scratch and use old skills as reference point?
      • Merge as it is (only consider major problems)
    • In future try to make skill integration faster (more incrementally)
    • Use parallel skill spaces for old skills
      • Not needed if incremental integration is done properly
      • Montreal showed how fallbacks to old code need a lot of work
    • Discuss responsibilities (Conveyor pose, new skills)
  • Conveyor Pose and Vision
    • merged into master already
    • Let Victor give the responsibility to Mario
    • combine conveyor pose and conveyor plane
      • initial estimate based on the plane
      • interfaces stay the same
      • adapt the plugins
      • let Mario do it if he has the time
    • dynamically make use of front and back lasers when they overlap
  • Navigation
    • Nick works on multiagent navigation
      • consider communication issues
      • current solution:
        • plan individually
        • share path (discrete points)
        • avoid collisions using simple priorities
        • on collision: drive to collision spot and wait until the higher priority bot passed
        • one could
      • local planning:
        • goal: move faster with less cpu consumption
        • use new approach using vector field histograms
        • similar to potential fields but with less downsides
    • releasing navigation:
      • release only parts of it will make release obsolete as it is
      • release a version with colli
  • Gripper
    • ready for testing
      • merge old skills
      • adapt camera transforms (camera needs different transform than the gripper, publish transforms for each axis relative to its mount)
    • open/close with rotating disk
    • WP detection with laser
      • same as used by mps in conveyor detection
      • exact mount position is still WIP
    • Feedback if gripper is open
    • z axis are flipped
    • camera is not moving with the z axis
    • springs are ordered but still pending (other springs work just fine)
    • Needs more mousepads (actually probably not)
    • try the smaller motors on 4th axis (improvement)
  • GRBL
    • looks promising
    • Axis movement using RapidMovement? seems to work nicely
    • Problems with the endstops
      • if one endpoint is reached all axis stop moving
      • caused by the gribble software
    • 4th motor not usable yet
      • Dirty branch from Daniel allows to test the motor with the old arduino plugin
      • clean that branch
  • Other hardware
    • change mountings of lasers
      • less screws
      • move laser closer to the center
    • Problems with laser calibration
  • Update on Smart Factory
    • new RealSenses? and NUCs arrives
    • end of march everything should be delivered
    • employee to get the mps setup running
    • No Robotino 4, Festo cannot deliver them yet
    • Instead new Robotino 3 (now we have a total of 8 Robotinos)
    • (Maybe running field somewhere in april)
    • MPS Software based on Team Solidus as baseline
  • Team Availability
    • Nick: available (gone in april for 15.4-18.4)
    • Morian: available (start thesis in winter semester)
    • Sebastian: master thesis done in end may/start of june, no attendance in GO, wants to go to RoboCup, further work if work at FH continues
    • Mostafa: master thesis in progression, probably NOT attending in GO and RoboCup, help preparation
    • Tarik: master thesis in progression (proposal pending), probably attending in GO and RoboCup
    • Daniel: master thesis in progression (proposal pending), probably attending in GO and RoboCup
    • Till: is back
    • Mario: probably attending to GO and definitly RoboCup, thesis finished this month and continue work afterwards
    • David: available, after 20th of march fully available
    • Victor: available
  • IFU
    • probably out of the team
    • maybe remain as sponsors
    • might stop funding students to attend the events
    • should get definite information soon
  • Workflow
    • great success so far
    • consistantly open issues if you face problems
    • do not talk in #dev
    • notification overload
  • Hackathon
    • was aimed at end of march but collides with RWTH exams
    • postponed until november
    • make team internal hackathon before GO
  • Action Items
    • Check github issues
  • League Changes
    • Competitive Orders
      • Single C0 competitive order, points as usual
      • +10 points for first team to deliver
      • -10 points for second team to deliver
    • graphical tool from ErForce? (open challenge) might be in place
    • probably no storage station, because the mps got changed
    • Prepare message to delivery station changes
      • do not specify delivery gate
      • specify the order to deliver instead
    • Maintenance may change (2 mainenances per bot, second maintenance costs 5 points)
    • Final rules get released soon
    • buffering a cap gives 2 points
    • resetting a ring station results in deducting
  • Misc
    • Have a working solution for the new gripper before doing major skill reworks
    • Adapt existing skills (David and Sebastian)
    • Look into the Docker timeout issues
    • Update Laptops to F29
    • Use NUCs from the smart factory to replace the laptops
      • add NUC to the base to shift weight
      • wifi antenna to the top (maybe a bad idea)
      • add another power source
      • future work, focus on getting a running system first
    • new SSDs sponsored by Magazino
    • Bring up failures of C-CS1 in TC meeting, no hope to get it fixed in time
    • improve simulation, get elaborated failure simulations deploeyed
    • Team Shirts
      • Logos mostly there, Sick is missing
      • David manages the shirts
      • Red Polo Shirts
      • Update the Carologistics Logo
    • Kiran will pursue the technical challenge (busy in february)
    • Mario will pursue the open challenge (OpenCart?)