This page contains conclusions about the project and class.

On the day of the final run, communication between bots had still not been achieved. Jim Garlick believed that protocol was correct and the failure was in the wiring of a walkie-talkie circuit board to the bot.

Mike Fisher presented the only bot that could perform all the tasks required of it, minus communication. The bot started in the hive, denoted by black paper on the floor, and set out in search of nectar using a random algorithm. Within a few minutes the bot correctly detected a red light (nectar) hanging directly above it. The bot then panned about 10 degrees at a time to the left, taking measurements of the brightness of the polarized light. When then brightness level dropped, the bot knew it had passed the center, and panned 10 degrees back to the right and correctly took off for the west wall. However, since one of the limitations of the bots was that they could not go perfectly straight, the bot paused about every two feet and panned for the brightest point in the light, always taking off in that direction.

Knowing that it had encountered the wall by activation of touch sensors, the bot turned into a position againt the wall, facing northward. The bot backed up until it hit the south wall, recording the distance. From there, the bot had no trouble using its wall-following algorithm to return to the hive. There the bot parked, ready to communicate to other bots directions on how to find the nectar, although none were available.

After pushing the start button on the bot, it began to seek the nectar on its own as a demostration that it's directions were accurate. The bot headed off to the left until it found the south wall, followed it to the end and turned the corner and backed up until touching the wall. The bot traversed the remembered distance out to the middle of the wall. The bot then executed a backing-up turn that put facing the polarized light. This turn looked like the reverse of what it had done earlier to get aligned next to the wall, so logically it ended up in the right place. From there, the bot again did its panning exercise, passing the brightest point in the light by a 10 degrees, and then turning back 10 degrees. This time, the bot would move in reverse for a couple of feet and take more measurements. This continued until the bot would back almost perfect under the nectar light, stop suddenly and beep with delight.

On subsequent tests that morning with the same bot, as it was backing towards the nectar it would appear to heading the wrong direction and then a couple feet away from the nectar it would change course drastically, 30 or 45 degrees, and make a beeline for the nectar. The best explanation for this would be the presence of a reflection of polarized light from somewhere in the room, that as the bot got further away from the original source, would become a brighter spot of light. Whatever the cause, the bot successfully found the nectar with each run.

Final Quotes

"Hardware and code works perfectly when its just you and the bot. When the demonstration comes, everything falls apart... its almost as if the little critter has a private vendetta against its creator." --Alex Reeder

"...not bot flies; *RO*bots!" -- Mike Fisher

"Touch sensors are the only ones you can trust -- every other piece of hardware will fail. Touchy-feely 'bots will eventually rule the Earth, after all other 'bots have accidentally thrown themselves off cliffs because they didn't know any better." -- Ian Bicking

"If only the class hadn't met at 8 AM..." --Mark Stosberg

Page design by Chris Hardie and Mark Stosberg. Copyright 1997, all rights reserved. Last updated 12/08/97.