So now that the book is done and Jo and I are back from India, we've decided to make robots.

Let me back up. India: It's huge, crowded, messy, exciting, vibrant, colorful, and rich in many ways. I can't hope to do it justice so I'll leave it at that for now. I can't wait to go back.

Now, about the hoverbots. We've been wanting to make physical bots for some time but have been preoccupied with other work. Today, at Rich's direction, I ordered a TI MSP430 development board and programmer cable. We also bought an IR proximity sensor with a range of 80 cm (31") and a pair of IR motion sensors with an arc of about 60°.

The fantasy is to wire the microcontroller into the drive mechanisms of a remote control model vehicle, ideally a (non-US branded) blimp or a black hovercraft. The proximity sensor would be mounted on the front to keep the robot from colliding with things. The motion sensors would be mounted on either side at an angle and allow the robot to sense the motion of humans or (hopefully) other robots. I'm imagining possibly a red or IR signalling light as well. The sensors, as well as the original RC circuit, would be wired into the ADC ports on the microcontroller and programmed with handcrafted C code compiled on mspgcc, an Open Source GCC toolchain for the MSP430 family.

We want to see a bot follow a person around, or follow other bots around, and collectively explore a territory. A small amount of C code could muster some convincing flocking behavior. The MSP430F1121 microcontroller only has 2K of program space and 256 bytes of RAM, so we're not talking about massive intelligence here, but autonomous and collective behavior from cheap consumer robotics. Anything more and we'd want to wire it up to a Gumstix and run Linux.