Shop OBEX P1 Docs P2 Docs Learn Events
Fill the Big Brain - Page 4 — Parallax Forums

Fill the Big Brain

1246759

Comments

  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-01 07:36
    Revised Test Software

    Try this attached revised version b test software. The LED will remain on longer for testing, improving visibility and allowing wire to pin 24 adjustment and testing.
  • jonlink0jonlink0 Posts: 11
    edited 2011-01-02 16:34
    Hello humanoido,

    Happy New Year!

    I've posted my robot project in a forum that actually allows in-progress projects ):< ( http://www.darwinbots.com/Forum/index.php/topic,3512.0.html )

    So, this is relevant because I think the software I'm working on might interest you. It combines a cellular neural network very similar to Brian's brain controlled by a genetic algorithm to allow learning. Unfortunately, I cannot begin programming until February, and even so the programming will take months, but I DO have the basic structure of the program laid out, and you are more than welcome to base your software off of my design. Good luck!

    -Jon
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-03 05:13
    jonlink0 wrote: »
    Hello humanoido, Happy New Year! I've posted my robot project in a forum that actually allows in-progress projects ):< ( http://www.darwinbots.com/Forum/index.php/topic,3512.0.html ) So, this is relevant because I think the software I'm working on might interest you. It combines a cellular neural network very similar to Brian's brain controlled by a genetic algorithm to allow learning. Unfortunately, I cannot begin programming until February, and even so the programming will take months, but I DO have the basic structure of the program graphically laid out, and you are more than welcome to base your software off of my design. Good luck! -Jon
    Jon, Happy New Year! Thank you for posting here. Wow! This is some of the best work I've ever seen and what you're doing is quite fantastic. The sharing of these ideas will undoubtedly lead to something remarkable.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-03 05:20
    Software Test 2

    It became necessary to have a small program for testing the data light to ascertain the current draw and compare the LED on the Parallax Propeller Proto Board (PPPB) with another one on another pin (mounted on the breadboard).

    Attached is testing Spin software for three types of LEDs.
    1) On the Demo Board to test the software
    2) Surface mount LED on the PPPB
    3) LED on the breadboard

    Pins are listed in the code and remarks.
    The delay between on/off cycles was increased to allow for ammeter settle time between reads.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-03 05:37
    Boards on Boards
    This is the board attachment phase

    Boards on boards simply refer to the mounting of secondary boards on the PPPBs. This is a small solderless breadboard (SBB) for wiring the components for data transfer (various BUS), data lights, power rail, power connection, and any additional circuits. It's also useful for rapidly changing and morphing the wiring and to run various test circuits. The board is extremely small and is very minimal in its propagation of EMI/ RFI and can be situated close to the Propeller chips' pin array - another useful minimization. This may or may not allow overclocking, future tests will determine the viability of overclocking. Currently the boards handle 80MHz and 20MIPs per cog.

    Although there are 20 PPPBs, only 19 will receive the SBB. This is because the PPPB in the Brain Stem is a hybrid which must interface to a BASIC Stamp and already has a tiny SBB made up of 7 total pin connector arrays. It may take some time to attach all SBBs. They are attached directly beneath the PPPBs label "PropClip" and "Keyboard." Note the hybrid PPPB already has the expansion kit installed for attaching keyboard, VGA, and mouse, so in the remaining boards this position is available for adding SBB real estate.

    SBBs are attached with large pieces of rolled plastic boxing tape. SBBs can be reposition, relocated, and removed at any time. The adherence may last for years, depending on environment and tape qualities.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-03 05:55
    Power Testing

    * Modified Parallax Propeller Proto board (both regulators and LED disabled) draws 4mA idle. Power LED converted to data light draws 20mA when on. Another test confirmed 4mA modified board current. Unmodified, board draws 10mA. To modify, lift the middle leg of the 3.3 vdc (output leg) of the LM1086 CS-3.3 regulator.

    * Another test, mod proto board draws 5ma, 19ma with on board data light. Put Taiwan LED on pin 25 with 150 ohm resistor = 12ma. Draws 6ma with 2.2k ohm. (since board draws 5ma, the Taiwan LED is drawing only 1ma)

    * Power LED converted to data light still has a 241 ohm resistor on it which is too small a value and that is why you see 20ma current. This is not acceptable for the purpose of battery portable operations. Therefore another data LED will be added.

    * The China red LED is the same as the red Taiwan LED – very bright at only 1ma. The green China LED is barely visible. (using a 2.2K ohm resistor). The green Taiwan is also too dim. The yellow Taiwan LED is also the same/ too dim.

    * The secondary LED was tested with a 2.2K resistor and was bright. With a 4.7K, it was less bright but still usable. You would not want to go dimmer. Therefore, with the Prop's 3.3 volt pin, use a resistor from 2.2K to 4.7K range depending on desired brightness. Both tests showed the LED drawing 1ma.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-03 05:55
    Testing Conclusions

    * The decision is made to use on-PPPBs data lights with 4K7 resistors.
    * The on board data light will be kept for more infrequent use
    * Modified boards will be used throughout the Brain Stem, Brain Base and Brain Span
    * A red LED will be selected over green and yellow
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-03 08:52
    attachment.php?attachmentid=76887&d=1294073469
    This 241 resistor is insufficient for regulating the
    data LED due to power constraints imposed by
    robotic battery operation requirements.
    304 x 260 - 13K
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-03 08:55
    attachment.php?attachmentid=76888&d=1294073708
    Small breadboards are added to PPPBs.
    257 x 226 - 18K
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-03 08:58
    Building DIY Breadboards
    A DIY project for Brain boards

    You can make very affordable home-made breadboards using pin sockets.
    Solder across pins on the sockets solder points on the bottom of the
    Parallax Propeller Proto Board to make connections.

    DIY Breadboards can be made very small. A four socket row is very
    useful for connecting data LEDs, resistors, capacitors, power connections,
    extensions, ground, and various sensors. The weight is significantly less
    than the smallest commercial breadboard at the Parallax store. However,
    for the purpose of constructing the Brain Base and the Brain Span, the
    larger Parallax is ideal for more connection points and larger scale tests.

    attachment.php?attachmentid=76889&d=1294073845
    This is one board from the brain's Brain Stem
    with a very small solderless breadboard built
    up from standard cheap dual row pin sockets.

    Economical DIY breadboards cost only a few
    pennies.

    Default.aspxDefault.aspxDefault.aspxMakeThumbImage.aspx?ID=%2fPortals%2f0%2fImages%2fProd%2f7%2f700%2f700-00012-M.jpg&PORTALID=0&W=120&H=120
    Commercial breadboards can fill in for
    Brain Base and Brain Span apps.
    Parallax item code 700-00012 is
    currently $3.99.
    287 x 214 - 21K
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-03 09:00
    attachment.php?attachmentid=76890&d=1294074023
    Testing setup for LEDs using breadboards
    213 x 276 - 20K
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-03 09:03
    attachment.php?attachmentid=76891&d=1294074121
    Getting ready to add more breadboards
    to these PPPBs. There are 19 total plus
    one which has a home-made SBB.
    282 x 211 - 19K
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-03 09:05
    attachment.php?attachmentid=76892&d=1294074321
    Finally the last PPPB gets a
    solderless breadboard!
    206 x 257 - 18K
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-04 04:34
    We'll need some simple examples of learning algorithms that can begin with one Propeller and then have it applied to multiple processors.
  • jonlink0jonlink0 Posts: 11
    edited 2011-01-04 19:20
    Alas, now comes the hard part ;D

    If you are interested in the Brian's Brain Cellular Automaton, I would recommend starting with the Propeller Life object (http://obex.parallax.com/objects/141/) and simply modify the cell-changing algorithm from line 362 onwards. However, one slight complication is that the Brian's Brain CA requires 3 cell states, while Conway's game of life only uses 2, so you might have to do more than just change variable values. I'm getting a spinstamp soon, so I'll ultimately be able to upload tested code, but in the mean time all I can do is provide recommendations. :P

    Although the BB CA won't be able to learn, I've found that starting with simple test code is much easier to work with than diving headfirst into the main project. Once Brian's Brain is up and running, you might be able to modify connections to make it more closely emulate a biological brain. One benefit of neural networks and cellular automata is they are more easily parallelized than symbolic AI, so I think at least some sort of non-symbolic approach would benefit best from your supercomputer.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-05 02:44
    jonlink0 wrote: »
    Alas, now comes the hard part ;D If you are interested in the Brian's Brain Cellular Automaton, I would recommend starting with the Propeller Life object (http://obex.parallax.com/objects/141/) and simply modify the cell-changing algorithm from line 362 onwards. However, one slight complication is that the Brian's Brain CA requires 3 cell states, while Conway's game of life only uses 2, so you might have to do more than just change variable values. I'm getting a spinstamp soon, so I'll ultimately be able to upload tested code, but in the mean time all I can do is provide recommendations. :P Although the BB CA won't be able to learn, I've found that starting with simple test code is much easier to work with than diving headfirst into the main project. Once Brian's Brain is up and running, you might be able to modify connections to make it more closely emulate a biological brain. One benefit of neural networks and cellular automata is they are more easily parallelized than symbolic AI, so I think at least some sort of non-symbolic approach would benefit best from your supercomputer.
    Indeed! Good places to start! I have noted there are numerous people working on systems with the goal of self reasoning. While it's possible to created complex software with great decisions and a multiplicity of branches, it does not represent the kind of universal self autonomy and unique problem solving abilities that we're after (because a human programmed it with each instance of problem solving). The brain should be able to solve problems in which man never even thought about. Currently I am trying to decide if this should be sectioned and divided into fields, i.e. expert problem solving in one specific endeavor field, then add fields. OR if it's possible to create that one universal algorithm that applies to all fields for reasoning and problem solving. I like to think it's going to be an extremely simple formula like Einstein's E=MC^2 or even his time travel equation.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-05 03:02
    Brain Blob BUS
    The foundation of the Hybrid BBB

    I had previously and briefly mentioned the hybrid aspect of the interfacing BUS for the Brain and this has now evolved through various testing trial and error into something more usable. The hybrid BBB is based on the proponent idea in the AM Algorithm Machine's Hybrid Bus, although it used a tiny 6-bit parallel version. For example, a two wire serial full duplex interface can use one wire or two wires just by changing the software. That's the idea. So we just put on the parallel bus and tweek software for three interfaces, i.e. make an 8-bit parallel and recycle 1 or 2 wires for serial duplex modes. The Hybrid BBB is a nice little hobby invention for getting more use out of a simple machine and increasing programming flexibility and minimizing the number of wired processor pins. It will also allow a system to be up and running sooner and leave the more complicated configurations for later. No schematics at this time for the Brain but there is a direct-connect schematic for the AM.

    For example,
    the AM Machine Interface

    attachment.php?attachmentid=74434&stc=1&thumb=1&d=1287044823attachment.php?attachmentid=74435&stc=1&thumb=1&d=1287045021attachment.php?attachmentid=74412&stc=1&thumb=1&d=1286956090

    This versatile machine operates in four modes, which can be wired as single mode A, single mode B, or dual modes A & B, or simply C mode. Operate the AM using a single wire and serial PBASIC code commands or program the machine using the seven parallel port bits.

    A - Single Wire Serial P0
    B - Parallel Bus P3, P4, P5, P6, P7
    A + B Dual BUS, as seen above
    C - Parallel Bus Only P0, P3. P4, P5, P6, P7
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-05 03:12
    Green Brain Blob & Green Guts
    Approach to recycling hardware and software

    The previous post specified the Hybrid BBB. Recycling may represent a large portion of the Brain Blob. The Brain, no matter how we look at it, needs all the resources it can get. Things inside will need the ability to serve numerous multiple purposes. I can see the use of shape shifters that are recycled from one purpose into another. Pins may recycle from output to input and visa versa. Memory is recyclable with varying code that runs and deletes. Algorithms can created "erasable" subroutines. Axons cycle for use and reuse. Code is self writing. Transposition, transmutation, juxtaposition, morphing, blending and evolving are all aspects of greening up the brain. For a comparison example of morphing where space is recycled, see the 4D Morphing Computer (with CoProcessor).

    attachment.php?attachmentid=67583&stc=1&thumb=1&d=1284503710
    attachment.php?attachmentid=67342&stc=1&thumb=1&d=1284503702 attachment.php?attachmentid=68056&stc=1&thumb=1&d=1284503736 attachment.php?attachmentid=68291&stc=1&thumb=1&d=1284503751




  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-05 03:20
    Brain Cog Power Draw
    Managing current within a Brain Blob

    attachment.php?attachmentid=76967&d=1294234474
    Each of eight cogs draw power
    when in use


    From time to time we will review the power consumption aspects of the Brain. We need to know the aspects of all power draw in the Brain as things are leading towards the development of battery operation for mobile robots (as one app). Dr. Braino has pointed out, as seen in the Propeller Specification sheet, a cog draws something on the order of 1 uA per Mkz per core and inactive cores are 0 uA. So every time a cog is turned on, it will draw a small amount of extra power. It may not seem like much, but multiplied by hundreds, or many thousand of cogs, it will add up and every milli-amp is significant.
    190 x 131 - 4K
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-05 05:54
    Brain Work Sheet

    ( ) 1 Stamp + 1 Prop
    ( ) 1 Prop + 1 Prop
    ( ) 1 Prop Master + 2 Props Slaves
    ( ) Parts Acquisition
    ( ) Brain Stem
    ( ) Brain Base
    ( ) Brain Span
    ( ) Power Tests
    ( ) 1-Wire Serial
    ( ) 2-Wire Serial
    ( ) Parallel BUS
    ( ) Serial Code
    ( ) Cog Consumption
    ( ) Breadboard DIY
    ( ) Breadboard Commercial
    ( ) LED Resistor Pair Value
    ( ) LED Red Tests
    ( ) LED Yellow Tests
    ( ) LED Green Tests
    ( ) Testing LED Code
    ( ) Sketching
    ( ) Researching AI
    ( ) Researching Cognitive Learning
    ( ) Researching High Speed Data Transmission
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-06 00:58
    All About Brain Serial Interfacing
    Deciding the nature of various connecting interfaces

    When is comes to serial interfacing, there are numerous options to consider. I have categorized some of the possibilities and listed some options to consider.

    First, decide if the interface will use one pin or two pins. A two wire interface (2 pins per chip) can operate in a bi-directional full duplex mode (send and receive at the same time or staggered). It can also be configured to function with one wire out of the two as half duplex.

    One wire (1 pin per chip) functions in half duplex mode. It can either send or receive but not both at the same time. One wire interfacing has the advantage of simplicity, can be up and running in a timely fashion, and has low code overhead potentially needing only a single cog.

    Standard One-Pin Uni-Directional True/Inverted Mode
    Standard Two-Pin Bi-Directional True/Inverted Modes
    Same-Pin (Bi-Directional) Inverted Mode
    Same-Pin (Bi-Directional) True Mode

    attachment.php?attachmentid=76995&d=1294308509
    Various serial interfacing (requires the addition
    of protection resistors)


    The napkin sketch shows four connection diagrams, less the protection resistors on each Propeller. Not the pull up and pull down configurations with a 4K7 resistor in bi-directional true and inverted modes. One pin bi-directional mode in inverted mode and pull down was adopted by BSS (illustration number 4). For simplicity, reliability and familiar code, illustration 4 will be adapted first unless otherwise noted.
    288 x 208 - 10K
    287 x 210 - 10K
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-06 01:17
    Resistor Selection
    Propeller Protection & Pull Down Considerations

    When connecting prop to prop, a 2.2K ohm resistor is recommended to protect the Propeller chip pin. In the circuit at the bottom of the diagram posted above, what value is recommended and does the 4.7K ohm pull-down resistor need a lower value? At first glance, it looks like 2.2K ohm could be used throughout. That would give 4.4K ohms to Vss with each Propeller chip, close to the original recommended 4.7K ohms. Considerations: how will this be affected with 20 props? With 50 props?
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-06 01:23
    Regulator

    The sketch shows the 3.3-volt power regulator and its pinout. The PPPB was modified by lifting the 3.3 VDC leg. Below the regulator is the pinout of the nearest connector.

    attachment.php?attachmentid=76990&d=1294305793
    274 x 243 - 7K
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-06 01:33
    Barrel Power Connector
    Basic connections

    This is a handy pictorial that can be used to measure
    power on the unmodified PPPB. It shows the (+/-)
    polarity of the Barrel Connector and the positioning
    of the power LED under the slide switch. It also
    identifies the green LED and its resistor pair.

    attachment.php?attachmentid=76991&d=1294306401
    283 x 260 - 7K
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-06 01:42
    LED and Surrounding Territory
    Pictorial Map of LED on PPPB

    Use this sketch made with a microscope to identify
    the original power LED and its companion 241 ohm
    resistor. This is converted to a data light by modifying
    the 3.3-volt regulator as previously described and
    soldering a wire to hole A and leading it back to a
    Propeller pin.

    attachment.php?attachmentid=76992&d=1294306926
    A microscope map hand sketched shows the
    positioning details of the LED for conversion.
    286 x 161 - 7K
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-06 01:51
    LED Schematic
    Sketch for modifying the LED

    You'll need this schematic to understand the wiring.
    junction holes A and B that connect Vss ground to
    LED D1 and resistor R1.

    attachment.php?attachmentid=76993&d=1294307465
    Schematic for the LED power to data mod
    281 x 134 - 4K
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-06 01:59
    Connector Pictorial
    Identifying the Propeller pins on the added connector

    This 20-pin socket (two rows of 10) is soldered to
    the PPPB. Note the location next to the crystal.
    This connector has the most pins used as noted.

    attachment.php?attachmentid=76994&d=1294307932
    Solder a 20-pin connector next to the Propeller
    to bring out the pins in two rows of ten sockets.


    This connector has everything to operate the
    brain in its current evolving state. It can handle code
    loading into all Propellers, transmit and receive,
    data LED, power and ground.
    288 x 216 - 7K
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-06 02:15
    PPPB Sketchings & Schematics
    Useful data notes for hookup


    Shows orientation for pin connector sockets, power
    regulator identification, Vss and Vdd and original
    PPPB schematics.

    attachment.php?attachmentid=76996&d=1294308886
    339 x 523 - 25K
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-06 02:24
    Another Mod to the LED
    Secondary mod brings power consumption within reason


    I just had an idea to insert another resistor between the
    Propeller pin and lead-back wire from the data LED.
    This will change power consumption from a full 20ma
    down to something more reasonable. It could be
    accomplished with software only but no reason to
    encumber the precious memory. Trial and error will
    determine the best lighting versus resistor value.
    Stay tuned for results.
  • HumanoidoHumanoido Posts: 5,770
    edited 2011-01-06 06:03
    LED Second Mod
    To bring LED power draw into proper range


    This is the second modification to the power light
    converting it into a data light. As you remember,
    the converted LED was drawing 19 and 20mA in
    two performance tests which is totally unacceptable
    in a multi-propeller system of brain magnitude.

    The solution is to add a resistor between the LED's
    lead (at point A) and the Propeller pin. Additional
    tests will determine the subjective value of this
    resistor (RBn).

    attachment.php?attachmentid=76998&d=1294322413
    The second mod to the data LED adds a
    resistor to lower milliamp consumption.
    This is idea for multi-prop systems with
    accumulated power consumption.
    283 x 206 - 7K
This discussion has been closed.