aporia wrote:I have a distant relative whose phd project was an AI insect coded in assembly. seriously impressive, if lacking in real-world application.
Does your beast recognise its environment? I seem to recall one of the US universities putting their robotic image recognition source code into the public domain and always fancied linking it up to some Lego creatures (currently with the USB proof-of-concept waiting for birth). I once worked with an ex missile chappie who wrote image recognition software for the australian missile defence people: apparently it's actually quite hard to get it right but you only need to get it wrong once.
What's on the other end of the cable?
The long length of ribbon cable snaking from the header on the device leads back to a breadboard where the microcontroller sits running the servos. At this stage the platform is basically a test of the motor cortex. In total there'll be four microcontrollers controlling the various subsystems in the finished beastie, so its a rather distributed system.
The sensory cortex won't be doing straight image recognition. It'll be equipped with the ability to sense light, sound, and IR in various directions in real time. The algorithm for processing that data takes its cue from the way real neurons integrate data using an adaptive shift register structure. It has its own reflex actions, so the servo that turns the "head" will be under its own control rather than the control of a central "brain". It still sends its interpretation of what's going on around it to the main CPU, however. I did some tests earlier this year and placed the "head" in front of the TV. It's really eerie watching its attention being grabbed by what's happening on the screen.
Ye Gods, I'm hungover. Must sleep.