A Mobile Robot that Maps Naively But Plans Intelligently

We implemented Yeap’s computational theory of cognitive mapping [1], [2] on a mobile robot equipped only with sonar sensors and an odometer. As expected, the resulting maps are erroneous. We demonstrate how a network of such highly unreliable descriptions can be used to find one’s way in one’s environment. The answer lies in the use of powerful commonsense reasoning to interpret the “erroneous” map intelligently. Our experiment consists of having a robot to traverse a part of our office environment once and then made six attempts to return “home”. Four of the six attempts were successful.