I've just started helping a good friend of mine, Lindz Lawlor, with his amazing Electric Giraffe. You can read all about it on the homepage there, so I'll just describe it briefly: It's a 17 foot high mechanical walking giraffe, capable of carrying up to 5 people! It has a 1000 watt sound system, and 100s of LEDs & lights. I'm helping to program the MAKE controller that will be the robot's brain. Currently, Lindz is installing servos in the jaw and ears and I'm linking them to the controller to bring some life to the beast.

Death Club Severed have released their first music video, a covery of Marilyn Manson's Long Hard Road out of Hell

This project was done for a university course in computer graphics. All of the computer graphics were rendered using a ray tracer called GKRenderer, written by Russell Pinnington and Andrew Fisher. The models were made by Andrew Fisher, Russell Pinnington and Graham MacFarlane.

Special thanks go to the Reading University LARP Society for the battle scenes!

The project took about 4 months to complete, with most of the time being spent on coding.

Download here! Death Club Severed - Long Hard Road out of Hell. (Right click and choose "Save As...") Feel free to share it!. File Size: 24Mb. Requires the DivX codec.

If you liked the video, Sign the Guestbook!

More information on the rendering engine coming soon.

Visual Target Acquisition and Designation System - A project designed to track multiple moving targets using inexpensive video capture software, such as a webcam. This project is currently under development - expect more information soon.

Introduction Project Chicken Walker, currently being undertaken by 1337-Robotics, aims to produce a bipedal walking robot, controlled by a neural network. The project is currently in the research and early design phase, although coding has begun on a Genetic Algorithms server that will allow potential robot controllers to be evaluated in a network environment, greatly increasing processing speed. Project Log To monitor the project's development, see the Chicken Walker Project Log. Genetic Algorithm A genetic algorithm is a type of evolutionary algorithm in which many possible solutions to the presented problem are tried in parallel. The best solutions (those with the highest score as determined by the fitness function) are recombined and mutated, and placed in the next generation. In this way, solutions improve incrementally with each generation until a suitable solution is found. In the Chicken Walker project, a special type of GA, Species Adaptation Genetic Algorithms (SAGA) will be used. These allow the length of the genome (the string of information specifying the solution) to change, unlike normal GAs in which it is fixed from the start. Through this, open-ended evolution, unconfined by prior restraints, will be allowed to take place, and the size of the controller will be able to grow as its complexity increases.

Distributed Processing

The Genetic Algorithm used to create the robot's controller is complex and requres a lot of processing time before useful results are gained. This would take a long time on a single machine, so the work is going to be distrubuted over a Local Area Network (LAN). In this way, many possible solutions can be tested in parallel.

Neural Network

The Genetic Algorithms are being used to create an artificial neural network This is a roughly a simulated brain: instead of a single complex processing unit as in a computer, it is made up of a large number of much simpler units (neurones) running in parallel. A neural network was chosen as it should be robust in the face of noise. The advantage of a neural network is that it can generalise: it can produce a sensible output given an unexpected input. This should help the robot when for example it encounters rough terrain. In the Chicken Walker project, neurons are of middle-complexity, and may perform functions such as summation, thresholding, trigonometry, exponentials and oscillations. While this is not how biological brains work, it is hoped that this model will lead to both faster evolution and faster running in the actual robot.

The GA is used to specify the types of neurones, the connections between them and the weights of the connections. There is no 'training phase' as with a normal neural network, instead all the weights are fixed by the GA.

Physical Design

The robot is loosely based on the skeletal structure of a bird. Birds have inverted knee joints, meaning that their knee bends the opposite way to a human's. It is hoped that by using this design, 'compliance' will be added to the system. This means that rather than being totally rigid, the joints will be able to flex which will not only be more energy efficient, but should also help with shock absorption and rough terrain.

Physical Construction

Although no work has yet begun on actually building the robot, it is planned to be constructed out of sheet aluminium alloy, as this material gives the required strength and is lighter than the equivalent in steel.

Many different types of actuators were considered for this system, including motors, servos, air muscles, shape memory alloys, hydraulics and pneumatics. The final actuator to be used has not yet been decided, but it will probably be either electrical motors or servos.

Downloads

The 1337 MUZ1K 574710N was a combined hardware and software project, the aim of which was to make a PIC microcontroller play music.

• PIC 16F877 Microcontroller @ 8MHz
• Millennium Development Board
  • Power Supply
  • LCD + Driver IC
  • Oscillator
  • Matrix Keypad
  • Piezo Sounder
• DAC
• Audio Amplifier
• Speaker

The programme consisted of around 1500 lines of C code. Keypad, speaker and LCD drivers were written, as well as an application which allowed the user to play the station as a freestyle instrument, or play along to pre-programmed tunes and get a score indicating how well they did.

It was intended that inbuilt piezo sounder be used for audio output, but we decided that this was unsuitable as it only allowed square wave output. We instead constructed a weighted-resistor Digital to Analogue Converter and fed the output of that into a pre-bought audio amplifier, which was connected to a 10W speaker. This allowed us to produce complex waveforms at the output, although we were limited to sum-of-sinusoids because of the very small amount of memory the PIC has. The audio driver allowed a number of sinusoids of differing frequencies to be played at once, and hence multiple notes and even multiple tunes could be played. This culminated in a rendition of "Pilgrimage" by Nine Inch Nails with both a normal and a bass line. Unfortunately, the clock speed was too low to allow notes to be fine-tuned, but the output was still acceptable and gave a much more pleasing sound than the other square-wave based projects.