Newsgroups: comp.robotics
Subject: Research student needs a
From: steve.jones@mecheng.fullfeed.com (Steve Jones)
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!yeshua.marcam.com!usc!cs.utexas.edu!bcm!pendragon!fullfeed!mecheng!steve.jones
Distribution: world
Message-ID: <af.9292.33.0NAFB49F@mecheng.fullfeed.com>
References: <CxxM2x.5LE@zeno.fit.edu>
Date: Thu, 20 Oct 94 17:47:00 -0600
Organization: MechEng / American Society of Mechanical Engineers
Lines: 135

SC>eeds assistance
SC>Message-ID: <CxxM2x.5LE@zeno.fit.edu>
SC>Newsgroup: comp.robotics
SC>Organization: Rockledge High School

SC>I am a ninth grader from Melbourne, Florida and am doing a ISEF
SC>science project on autonomy applied to robotic structures and
SC>three-dimensional vision aided by neural nets and expert systems.

SC>NOTE: Please take !fingers! to include the thumb unless otherwise
SC>specified. 
SC>This short essay is in four sections: HELP!, Video Input/Processing,
SC>Robot Arm Mechanics, and Official Description - Please Comment.


SC>HELP!:

SC>. If you have ideas, had experience with any related experiments, or
SC>        actually work with these topics, please feel free to describe
SC>        your achievements to me.
SC>. For the program that extracts 3D information from the images, I!d
SC>        like advice on how to go about it (other than simple
SC>        explanations of how to triangulate the points! positions    I
SC>        mean deciding how the points match up between the views,
SC>etc.). . Suggestions on alternative setups for the !stereoscopic
SC>imager! and         arm/fingers would be welcomed.
SC>. I need advice on elements to be considered for each aspect.


SC>VIDEO INPUT/PROCESSING:

SC>To attain the stereoscopic images I!ll be working with,I have a color
SC>composite video digitization board and am planning to connect it to a
SC>color video camera, with four (relatively) high quality mirrors
SC>mounted before it such that each vertical half of the CCD gets a view
SC>of the scene from a left or right perspective respectively:

SC>?
SC>   *    - object


SC>\\//    - mirrors
SC> ~~     - video camera
SC>TOP VIEW

SC>I don!t know whether I want the mirrors to be automatically
SC>positionable, or even manually for that matter.  I!m also questioning
SC>sensors on the mirrors to locate their position and orientation, or
SC>just to calibrate on some zero.  Anyway, that will give me a
SC>stereoscopic image to work  with. The software will then need to
SC>identify vertices in the image.  To do this a loop will run through
SC>each scanline, placing a marker whenever the average intensity of
SC>surrounding pixels changes by more than some threshold.  The lateral
SC>disparity between each pair of control points will then be computed,
SC>and the scene will be translated into a three-dimensional map.  Then
SC>the classification system will fire up and attempt to break the object(s
SC>) down into basic components and recall and/or learn and store their
SC>identity. 

SC>ROBOT ARM MECHANICS:

SC>As stated later, I!m building a robot arm with five degrees of
SC>freedom and an index finger and two opposing thumbs.  The thumbs will
SC>be like the fingers, basically a one- or two-jointed fingers mounted
SC>on a rotating base.  The arm will have SHOULDER ROTATE, SHOULDER
SC>HINGE, ELBOW, WRIST ROLL and WRIST YAW, respectively, down its
SC>length.  Basic setup is shown below: ?
SC>|  _ _
SC>||O_O_)
SC>| \  `JOINTS
SC>   ROTATING BASE
SC>THUMB
SC>   \|/ - HAND
SC>  _O_ - WRIST YAW
SC>  ___ - WRIST ROLL
SC> //
SC>  O -ELBOW
SC>   || 
SC>__O__ - SHOULDER HINGE
SC>_____ - SHOULDER ROTATE

SC>The arm!s motors will be steppers from old 10-, 8-, 5 1/2-, and
SC>3-inch disk drives.
SC>I am contemplating using an old 8085 chip to receive the computer!s
SC>commands and take over the task of actually doling out drive signals
SC>to the arm!s individual motors and buffering the arm!s sensors!
SC>responses.  Right now I!m planning on having only one zero position
SC>detector at each joint, and perhaps for the fingers (and thumb) one-
SC>or two-stage lightweight pressure sensors, and maybe an optical
SC>detector pair between each finger. 

SC>OFFICIAL DESCRIPTION - PLEASE COMMENT:

SC>        Can a robotic arm be constructed and linked to a computer
SC>with a stereoscopic imaging system and be taught to classify
SC>three-dimensional objects, and if so, what factors affect efficient
SC>architecture?         If a standard industrial/experimental type
SC>robot arm, with five degrees of freedom an an index finger and two
SC>opposing thumbs, is constructed, and a video camera, combined with a
SC>image splitter and connected to a computer with appropriate image
SC>processing and neural network/expert system software, is compiled,
SC>then the system will be taught to classify three-dimensional objects,
SC>and the factors affecting the efficiency of the Architecture can be
SC>assessed and quantified. 
SC>          1) Have human subjects create objects for training
SC>repeat as necessary.
SC>          2) Create stereoscopic imaging system.
SC>          3) Create software vision interface.
SC>          4) Create stereo disparity computation filter.
SC>          5) Create neural network/expert system to classify objects.
SC>          6) Create arm.
SC>          7) Create software arm interface.
SC>          8) Create arm controller neural net.
SC>          9) Create master controller to link each subsystem.
SC>        10) Train each net repeatedly, experimenting with
SC>manipulation of each network!s controlling factors     i.e. number of
SC>neurons, layer layout, back propagation factor, learning speed, etc.
SC>        11) Employ human subjects in the evaluation of the systems!
SC>responses! accuracy.
SC>        12) Repeat steps 10 and 11 as necessary.

SC>Please send responses to my instructor via e-mail:
SC>scampbel@rhs.brevard.k12.fl.us

SC>Thank you for your assistance.
SC>Chris Campbell
SC>Rockledge High School
SC>Rockledge, FL  USA

This is a ninth grade project!!!!   Man!  Makes my high school education look 
like pre-school finger painting time!!  Good luck in the project.

---
 * CmpQwk #UNREG * UNREGISTERED EVALUATION COPY

