{"id":219,"date":"2019-02-22T02:04:14","date_gmt":"2019-02-22T02:04:14","guid":{"rendered":"http:\/\/mrsdprojects.ri.cmu.edu\/2018teamc\/?page_id=219"},"modified":"2019-02-22T02:47:51","modified_gmt":"2019-02-22T02:47:51","slug":"system-design","status":"publish","type":"page","link":"https:\/\/mrsdprojects.ri.cmu.edu\/2018teamc\/system-design\/","title":{"rendered":"System Design"},"content":{"rendered":"
\nSystem Requirements:<\/u><\/span>
\n
\nMandatory Performance Requirement<\/span><\/p>\n
    \n
  • M.P.1 Will operate at speeds up to 5 m\/s<\/li>\n
  • M.P.2 Will have a positional precision of 2m<\/li>\n
  • M.P.3 Will operate at maximum wind speeds of 5km\/h<\/li>\n
  • M.P.4 Will achieve a detection accuracy of 70%<\/li>\n
  • M.P.5 Will achieve classification accuracy of 80%<\/li>\n
  • M.P.6 Will have a grasp success rate of 30%<\/li>\n
  • M.P.7 Will grasp blocks of 1.5kg<\/li>\n
  • M.P.8 will grasp blocks of 1.2 x 0.2 x 0.2m<\/li>\n
  • M.P.9 Will detect poor grasp with 70% success rate<\/li>\n
  • M.P.10 Will transport blocks with a 70% success rate<\/li>\n
  • M.P.11 Will place blocks with precision of 0.1m<\/li>\n
  • M.P.12 Will assemble wall of 3 layers<\/li>\n
  • M.P.13 Will detect poor placement with a 50% success rate.<\/li>\n<\/ul>\n

    Non Mandatory Performance Requirement<\/span><\/p>\n

      \n
    • M.N.1 Will have a manual overwrite safety feature<\/li>\n
    • M.N.2 Will be easy to operate<\/li>\n<\/ul>\n

      \nFunctional Architecture:<\/u><\/span>
      \n\"\"<\/p>\n

      Cyberphysical Architecture<\/u><\/span>
      \n\"\"<\/p>\n

      System Description:<\/u><\/span>
      \n
      \n<\/p>\n

      \"\"
      \nOur system uses a hexacopter platform with manipulator attached to pick up
      \nblocks for construction purposes.<\/p>\n

      1. External Base Station<\/span><\/p>\n

      The external base-station consists of an workstation (preferably a laptop, because
      \nit is portable), that is capable of sending commands or simple instructions to the
      \nhexcopter platform, and receiving feedback messages. Here, we use the workstation
      \nto analyze to further analyze a structure provided by the user. The workstation
      \nwill work on creating a pick-up routine or schedule based on best order picking and
      \nplacement.<\/p>\n

      2.Perception<\/span><\/p>\n

      2.1 Block Detection<\/span>
      \nWe employ a set of algorithms that enable us to detect blocks from both a farther
      \nrange (as a part of scene understanding), as well as at close range. We use basic
      \nobject priors such as colors and depth regions to reduce the search space, and pass
      \nthese regions through a low computation object detection network pretrained to
      \ndetect blocks in simulation and create an initial set of bounding boxes around the
      \nblocks. The goal is to reach within 2m accuracy of the point from where all the
      \nblocks are visible so as to ensure reliable servoing.<\/p>\n

      2.2 Block Identification System<\/span>
      \nThe block isolation system, or the pick-best-block algorithm is mainly used to pick
      \nthe best block given a candidate set of bounding boxes. Here, we will use the fusion
      \nof visual information from different sources to pick:<\/p>\n