A new system of robotic navigation being developed at Purdue University enhances a bot's ability to get around unfamiliar environments by allowing it to "guess" what unknown areas look like. Unlike the traditional method of "simulataneous localization and mapping" (hilariously nicknamed SLAM), in which bots take detailed measurements of an area to generate maps, the guessbots divide an area into cells and make predictions about "frontier cells," or areas adjacent to cells which have already been mapped. Each frontier cell is then assigned a "confidence score" -- cells with low scores need further investigation, while those with high scores can be added to the map. Initial computer simulations of the system resulted in virtual guessbots needing to map less than 33% of a building to navigate successfully, and while real-life prototyping already underway doesn't seem to have gotten quite as good, it's produced higher-quality maps in shorter amounts of time than SLAM, according to the developers. The guessbots do have limitations, however: they system only works well in highly-structured environments like buildings -- outdoors is probably out -- and like all robotic navigation systems, small measurement errors add up fast. Even still, let's hope this filters down to the consumer level fast -- a navigation system that's able to predict what's coming up next instead of blindly following a map just might keep us from wrecking all those cars.