PAGES

Saturday, August 18, 2012

Giant RX Propeller Super Computing Brain Robot

The New Giant RX  (c) humanoido
Giant RX
Propeller Super Computing Brain Robot

The logical evolution from the RX Retro Super Computer

5 out of 26 boards shown
Introduction  The RX Propeller Supercomputer is a Delineated Super Computing Machine robot built at Humanoido Labs. The Giant RX Propeller Super Computing Brain Robot is the next generation. I started putting together multiple Parallax processors in 2002 to make humanoid brains but by the year 2010 the robot brains had arms, airports (a space program) and commanded flight crafts remotely and autonomously.

Enhancements  The new RX is evolved to focus on enhancements inside the machine and perpetuate the hardware designs before insertion into a humanoid head. The enhancements investigated include overclocking, super cooling, rapid acceleration, distributed parallel, cloning, programming, threading, processor enhancement and derivation, and introduction of new concepts such as the TMACH, a machine that can transform inside the RX.


Specifications Overview  Ten years later and based on Big Brain Technology, the new RX is DISP configured with 800 Propeller RISC computers, 3,200 IO Ports, 100,000 threads and interface ready for more humanoid robot peripherals and sensors. The next delineation goal is reduced mass to fit humanoid heads about the size of Honda’s ASIMO.

NTSC TV inserted into the PPPB belly of the robot brain for mobility protection.

Brass standoffs allow the connection of additional processor host boards for great platform compatibility. Additional ports sieed at lower right connect a mouse and keyboard. Lower left shows a large 4 line Serial LCD used for programming and debugging.

Build  A portion of the assembly is formed by twenty Parallax Propeller Protoboards PPPBs, several host boards, and four giant boards which are fabricated from a number of solderless breadboards. As seen in the photos, there are at least 26 boards holding a mix of Propeller processors and the hosts. The hosts include other processor variants that open the door to multiple platform compatibility. The PPPBs are modified to reduce power draw across the large array. The voltage regulator is lifted up from the board and the LED is rewired. It forms a partial structure which is wired on the outside surface with added small solderless breadboards. The giant boards set in a cage and connect with jumpers and can hold hundreds of Propeller chips. This machine is readily capped at around a hundred Props. The third Partition holds 50 more props but can operate at any number less than 50. This creates a flexible machine with versions to function anywhere between 100 and 150 so I can pull off chips for other projects, testing, and temp setups.

A robot fist for security
The Robot Fist
As an experimental simple arm for testing the condition of many output ports and for the use as a possible security device, a small punching boxing glove fist was introduced. To keep things simple, the fist has a wooden arm that does not bend, but rather lowers with some force. To activate the lowering method, a servo trips the mechanical switch releasing the cocked fist.

http://humanoidolabs.blogspot.tw/2012/09/robot-fist.html

Robot Classification
This interesting project has been identified as a robot, robot brain, and robot super computing machine. It drives a robot fist currently which is servo released from its cocked position to punch, as a test for any of the I/O ports. The older retro photo does not show the fist. When time permits, a movie will demonstrate its action.


Build Part 2  Four Giant Boards wired as sections in Parallel help make up the three Partitions of 50 props to each Partition, running with ParP. Programs load into each Cog inside each chip. Chips contain 8 Cogs, more cores result with enhancements. Plug pin cables route robotic peripherals as needed such as the Big Red Fist and Blue Little Fingers. If needed, the machine also runs with resources from another computer including the monitor, keyboard and TeraByte Drives. Generally the RX is run self contained. Parallax also has the free PST for programming and debugging. The two most popular languages include Propeller Assembler (for highest speed ratings) and Spin (for ease of programming in a high level language).

Design  The machine is a massively parallel design with three PARPs Partitions that hold 50 Propeller chips each (per description) and operate in Parallel-Parallel. The input is with a connected PC, or separate keyboard, mouse, and sensors. Output is to a TV, LCD, LEDs, other computers and various devices. The RX can also speak and sing with a plug in speaker amp and the posted programs. It will even run existing serial communication programs with a little tuning for ultra high speed serial.


Computer Programming Languages  The robot can run from a selection of around 230 programming languages and their variations but I've only had time to explore a handful. The language also determines the speed and function type of the machine, with Propeller Assembly being the fastest.

Running GCC provides full floating point while Spin is integer based. The machine can also program in C, C++, SPIN, BASIC, FORTH, and numerous others with options of Float Compute or INT based. I've added some of my own languages, like OPTICS for use with Optoelectronic Interfaces, DISP for Distributed Parallel, and LITTLE BLUE FINGERS for internal signing. DISP is the result of the parallel algorithms developed over several years.

Wiring  The Giant RX Propeller Supercomputer Robotic Brain has two wiring platforms, one is with SW and the other is with HW. SWW is a relatively new concept for the Big Brain and is now handed on to each new machine. SWW adds capabilities of Transforming and the creation of TMACHs.

Algorithm  The machine is testing the boundaries of Distributed Parallel. The DISP language is written and distro delivers the code. The machine is designed to take an application for solving and distribute it across the machine's processors for simultaneous processing. Sub solutions are garnered and applied to the final conclusion. Algorithms compute in binary and hexadecimal or release results converted to base ten and strings.

Software  The programming software is built up over ten years. It include DISP for programming in Distributed Parallel, a Neural Loader that places code into all processors, The SW Enumerator that quantifies each and every processor, RTOS real time operating system that keeps track of operations, timing, processors, threads etc., a Cloner for cloning software circuits, a VIPS Builder to enhance the Propeller chip, and new software is being introduced to create additional true parallel Cogs to increase the number of deterministic cores in each chip.

EXO Tower Config  (c) humanoido
The 1st config EXO Tower began first as a formed rectangular exoskeleton made up of 20 PPPBs and 160 RISC computers from Parallax. All four sides contain processor boards. The belly is reserved for TV. One end is for connecting the mouse, keyboard, TV and displays LED arrays. Each printed circuit board has an attached miniature solderless breadboard. This assists rapid wiring and rewiring. Note the parallel wiring beginning to take shape, along with the power supply circuit mods and the LED changes made to create a lower power consumptive circuit.

Power Supply  The entire machine runs off a Chinese variable voltage Lab Bench Power Supply. This allows variance centered around the Propellers 3.1 optimal voltage in standard mode and gives real time digital readout indications for voltage and amperage. This works ideal for experiments with low voltage, acceleration, super cooling, hypering, overclocking, and various enhancements to tune settings. The Bench Lab Supply can be replaced with batteries in the low power version for portable and auto operations. A lower amperage draw condition is forced by special programming of cogs and pins.

Open Cage  The open cage concept works well for access, rewiring, modifications, tuning, observing the LED outputs, and the hookup and attachment of plug 'n' play peripherals. Additionally, it facilitates the mounting and operations of the super cooled chamber machine and pressurization canisters, to enhance and accelerate chips well beyond their standard speeds.

Robotics  The brain is heavily robotic with interfacing to servos and numerous sensors. For example, it borrows the Robot Fist technology from the Big Brain, a big red punching Fist that was designed so the brain can protect itself and punch out negativity. The RX is designed for the new robotic internal signing machine which originally came from Small Blue Robot Fingers, and it can also do transforming. Small Blue Fingers has its own language that is transmitted through Transparent Plastic. Keep in mind, the robotic interface, and things with robotic actions, also take place internally inside the RX brain. A pin can create at least three states, wiring itself autonomously and thus becoming a finite state machine. The RX can produce up to one quadrillion of these states for various robotic explorations. For this type of mobility action, no servos are necessary. The other robotics that made up the RX plug 'n' play aspects include: servo control, Ping, PIR, Transducers, IR Transmitter, IR receiver, Satellite GPS Sensors, Atmospheric Pressure Sensors, Gravity G-Force Detectors, Scanning Platform, Wireless Radios, Remote Control, Compliment Security Arm, Helicopter, VTOL Experimental Quadcopter, Giant Optical Eye, etc. etc.

Transforming  The RX has led to the Transformation Machine TMACH. Briefly, this allows the RX to achieve many more variants, including multiple machines inside machines, and new types of machines called MIMs or Machines Inside Machines. TMACHS also permit the RX to transform into many different machines, different style MIMs, link up together or go their separate ways. One aspect we are beginning to explore is the use of Internalized TMACH Swarming Mims or the TMACHSMIM. This facilitates a FLOATED object that can transcend/move from location to location instantly within a machine or from machines to machines, as Teleportation. This type of Teleport is thought to speed up the machine significantly.

Machine Overview  I view the design and operation of the new RX machine to serve as a robotic brain for advanced humanoid robots, scale down to make power brains for hobby robots, and to serve as an interesting way to intelligently control the outside world, plus a very important aspect to explore the internal microcosm of the Propeller chip array, learning new things, and go to explore new processing worlds.

Presentation Photos
Pics are yellow-gold sepia tone to give a retro 1950s look because the project was originally the RX Retro Supercomputer. I have chosen orientations that best represent and highlight various sections. Pics are presented in a personal artistic rendition. Pics show wide views of the machine. For closeup views of the chip, see this link or below.

Propeller chips. Photo courtesy Parallax Inc.



Use of Materials & Challenges
This is a low budget personal super computing robot project using my own money as a hobbyist and working alone in my spare time. After investing in processors and boards, I did not have resources for fancy packaging so I chose common inexpensive materials that were available locally at budget cost. Parts were found at hardware stores, dollar stores, and grocery stores.

Design & Size
I like to resort to interesting designs and invent new things with new large machines. This science fact creation is fully operational. This project is a super computing robot so the size is large in a towering configuration. To compress the machine into a robot head, in a future project, it’s likely boards will not be used.


Future Project Projections
Humanoid Head Containment

The next incremental project may have an approximation-polymer-humanoid-head made with the parts custom fit inside.

Physical Components
SMDs won’t unplug for this config so the project will stick to DI Processors. To lighten and compress the load, no boards will be used. This is an adaptation spinoff of the Sphere Computer previously posted. In the Sphere Computer, processors layered the curvature of the sphere internally and optically communicated with their neighbors in a growing centric density of spheres defined by their calculus.

Three Dimensions
A new array design is in the works to get everything to fit by changing from the 2D boards seen in the photo to 3D geometrical constructs with at least an XYZ component. Spheres may not be used as realistic humanoid heads are rarely round. There is some idea floating around about including a novel idea internal creature with several DOF and mobility or mechanical signaling. It would take two micron servos to accomplish this and could demonstrate autonomous repair, reconfiguration, upgrading, add-on, or rebirth of internalized structures. For example, a hand could re-array a processor chip and bump it up or remove a suspect, or add new ones from a supply cache, thus negating intervention by humans in to robotic brains.

Hypering & Wiring
This one will probably not engage a Hypercube as past projects have shown such a design requires an inordinate number of vertices connected. New design technologies are leaving the hard wiring behind so it's hard to tell what new wiring will reveal this year and next year. A new kind of robotic brain internalized wiring with software is permitted.

Design & Chip Selection
Intentions are to maintain the parallel constructs, DISP design, algorithm and language. I will definitely continue to use the Propeller chip. I learned this chip has additional bonus “core processors” inside (beyond the eight listed on the specs) which can be used for complimentary super computing task management in my specific project. Expect a write up on this.

More Chip Enhancements & Ratings
Hot rodding the chip can happen several different additive ways making the flexibility of this chip the right stuff for the project. In my opinion, the specs on this chip are under rated because it can hyper to a relative speed around 10 to 20 times greater. Parallax doesn’t tell you that because their job is to put out this new chip and provide excellent fundamental details about it, however, it’s up to you to do thinking out of the box and give the chip new functionality.

Programming
Many of these projects are about that new functionality. Also keep in mind that many programmers learned programming in one fundamental way and are not likely receptive to radically new designs and are not looking for things out of the box. One example is the relatively few number of people taking advantage of the solid performance hit available with overclocking and super cooling. Even in new brain super computing designs of parallal PARPs, Distributed Parallel, Cloning, or Neural Diversity, for example, and few people, if any, are grabbing these golden opportunities. One purpose of this ongoing project series is to introduce these concepts for hobbyists and filter the technology to stimulate the creation of more intellectual robots that can do levels of machine thinking with a degree of awareness previously unknown. That’s why so much time was spent developing DISP. The programming language/Algorithm Suite is suited specifically for the Distributed Parallel Computing Machine and uses the algorithms that were developed on the AM Algorithm Machine and based on work since 2002. Parallel Computing with multi core chips is absolutely mind boggling when one experiences the magnitude of screaming power and capabilities (relative subjective descriptive terms).

Project Time Frame & DIY
Some of these projects are big - based on thousands of hours of work, and some were built over a period of years. A specific number of quoted days is only relative. That’s because one machine builds upon its predecessor. Adding up the work of predecessor machines is a more adequate determination. With patience in using the general concepts, it’s possible to make smaller, yet high powered, low cost hobby variants, for example, with 96 cogs instead of 1,200 cogs and 12,000 cores instead of 150,000 cores. This will bring down price, assembly time, and complexity.


Another Design
There is another design in the works for a very low cost hobby robot brain that gives your projects the ability to learn what  you teach it in English. This may be a small one board solution. Coming soon!