|Cin-D LOU House Bot|
|Start Date: 1/1/2012|
December 2012 Update: Members Jon and Craig stand in to show how Cin-D LOU's Kinect 3D vision is able to perform real-time recognition and construction of their skeleton joint positions.
There has been a lot of behind-the-scenes work happening on the Cin-D LOU project during the past months.
Jim got a copy of the MAKE book "Making Things See" by Greg Borenstein. If you don't know the story, it is better told in the book.
The Kinect is a Microsoft product; a peripheral originally developed exclusively for the Xbox 360 video game system. This amazing USB-connected infrared camera provides a distance/depth image for objects placed in its field-of-view. It is a result of many years of academic research in the field of computer vision, with hardware development by the Israeli company, PrimeSense, working in close cooperation with MicroSoft software and algorithms. The "Microsoft Kinect" product was launched on November 4, 2010, and was a major commercial success. It sold upward of 10 million units in the first month after its release, making it the fastest selling computer peripheral in history.
On the same day as the Kinect's release, the NYC open-source hardware company Adafruit announced a $2,000 bounty to the first person to produce open source drivers that would let anyone access the Kinect's data. Wehn MicroSoft reacted negatively to this bounty, Adafruit increased the bounty to $3,000. Six days later, on November 10, Hector Martin claimed the Adafruit bounty with the first public version of a working driver. He then joined the OpenKinect project which continues to improve and maintain the drivers to this day. HIVE member Craig brought a Kinect http://wiki.hive13.org/Kinect
programmer Josh Blake
The story of the maker community's hack of the
See the new page about making the eyelid frame parts http://wiki.hive13.org/Cin-D_LOU_EyeLids.
Cin-D LOU is the HIVE's house bot project; an autonomous robot (at 5'-2" tall) soon(?) to be able to self-navigate around the HIVE space with some initial (and ever growing?) ability for interaction. This project is our version of a home-built C3PO-like protocol droid.
We are using a motorized wheelchair as the modular base unit. The inital "skin" has a nominal female form, but can be changed out to have a male or androgynous form.
Our project team is an informal group of like-minded folks. We hold our weekly team project meeting on Tuesdays, before/after the regular business meeting and then make progress each week. Individuals and small groups can take on specific tasks and interested bystanders can follow the progress and contribute as things evolve. The purpose is to learn and have fun!
The project has four initial areas of effort.
(1) Locomotion - Thanks to Jon, we now have a slightly used, model 1122 Jazzy motorized wheelchair from Pride Mobility Products. http://www.pridemobility.com/resourcecenter/Downloads/Product_Owners_Manuals/product_owners_manuals.asp The base came in with years of use and is fairly heavy with batteries and power circuits. Different control schemes to recognize position, do path planning and execution are being determined. The HIVE's MS Xbox Kinect might be one input, plus others that are TBD.
Above are some photos taken while stripping down and cleaning up the motorized wheelchair base. Formula 409, WD-40 and elbow grease do wonders on grime, rust, and nasty hair balls wrapped around axles. It has been re-assembled in stripped-down form and is ready for further hacking the safety bumpers, line-tracking navigation parts, and everything else.
The project team is currently looking into different schemes to tap into the joystick and/or powerboard to get computer control for the drive motions. As always, there are others out on the web that are already doing similar things:
Once we get computer control of the motion, our first thought is to start with a simple line-following strategy like one of the following links:
We want to quickly graduate to autonomous navigation in an open space and are currently use of the open source Willow Garage Robot Operating System (ROS) capabilities.
(2) Arm and Hand - Here we're thinking to use laser cut acrylic for the fingers, hand, wrist, forearm, elbow, upper arm and shoulder joints. Planned actuations come from forearm mounted small DC motors with leadscrews driving tendons. Stepping motor drives with gear reduction are anticipated at the elbow and shoulder. The ability to execute programmed motion sequences like pointing, waving, presenting a drink, making a fist bump or shaking hands without (or with) feedback are TBD using networked Arduinos or such for control.
(3) Head - Here we've got a styrofoam head with a vacu-formed replaceable female face as a possible starting point. There are some initial mechanical designs for eye balls (with blue LED pupils) that track with up/down left/right motion and working eyelids and eyebrows. http://www.youtube.com/watch?v=uYYqycOWH5g Mouth and speech capabilities to be determined. Neck functions to turn and nod the head would be included. Again, we anticipate the ability to execute programmed motion sequences (nodding up/down for yes, shaking head left/right for no, and eye and head movements during motion tracking) without (or with) feedback. Here is the start http://wiki.hive13.org/Cin-D_LOU_head
(4) Body - Here we're looking for the HIVE artist types to make a simple light and quick body space frame from exacto-knife cut or laser cut cardboard or other layers using the AutoCAD 123D Make capabilities at http://www.123dapp.com/make. This would get us an egg crate or open grid structure that would have the outer body profile with cavities inside for mounting necessary controls and conduit paths.
Finally, if you have read this far, here is one of many inspirational robot sites out on the web: