The mobile scout served to calibrate and retrieve projectiles for the immobile launcher. A spiral grid was used to simplify the problem of map registration so the scout could accurately report back to the launcher where a projectile had landed. Complicated interactions between the two robots and tricky localization problems made for a challenging and successful project.
Walter, the pool-playing robot, could distinguish between striped and solid balls and then put them in the right "pocket" of the pool table. It found the balls using IRs (which it shut off to conserve power when the IRs were not needed) and then used the Lego active light sensor to distinguish striped from solid (by holding the ball in its gripper and rolling the ball so that the ball's surface rotated in front of the light sensor).
Interactive Robot that used emotional states (tied to PID controllers) to exhibit interesting behaviors. Emotional "highs" decayed with time and various "hungers" would rise in urgency if not satisfied. The robot had sensors to detect being petted, food (colored patches on the rug), and light. It had effectors to move about, whirl to exhibit excitement, a tail that could exhibit a range of internal states, and a wide range of sounds correlated with internal states.
Matt and Paula started talking about this project during the first month of class. The robot was a realy a chess-piece-recognition-and-moving machine. The robot used IRs to communicate with a Sun that ran GnuChess and that generated the moves. Good design coupled with lots of hard work made this project successful. Note that dental floss makes a pretty good source of "rope" for use with lego pullies.
Used information printed on a board to detect gradients and localization hardware to learn a map and use it to navigate. The pieces worked remarkably well but even with removing all extraneous code from the legOS kernel they couldn't fit everything into the available memory.
Used a grid painted on a surface to reduce localization and path following to recognizing and counting grid lines. Doing this robustly turned out to be more difficult than originally thought, but robust error correcting codes resulted in interesting if not perfect performance.
Ths was another project that relied on accurate localization and fell somewhat short of its initial goals because of this reliance. They also spent a lot of time working on their electromagnetic gripper which was problematic mechanically and electrically and a time sink for their project.
Required recognizing cans and using a gripper to pick them up. The full project also depended on a reliable method of locating and returning the cans to a recycling center for disposal, but this part of the project was never successfully realized.
Required orchestrating complicated behaviors. This project also involved building and integrating new hardware, a sound sensor in this case. The sound sensor was meant to enable the robot to respond to "voices" but in the end it was only effective in detecting hand claps and even then not terribly well.
A pretty ambitious project that started out in NQC and then had to be ported to legOS when memory became a problem. Their final demonstration was pretty impressive.