Whistling in the dark

Abstract

We demonstrate that a simulated group of robots can co- operate to robustly transport resource between two areas in an unknown environment using an algorithm inspired by the trail following of ants and the waggle dance of honey bees. Rather than directly marking their environment, the robots announce their successful paths through a common localization space. It is found that the algorithm is robust to significant localization error, suggesting that the method will be viable for teams of real robots. relation between the representations maintained by two or more individuals. A prime example is the Global Position- ing System (GPS). Two systems equipped with GPS share a metric localization space in planetary coordinates. Simi- laxly two robots that start out with known positions in the same coordinate system and maintain a position estimate via odometry share a localization space. In both examples each robot has only an estimate of its position in the true space, but the true space is common to both. More abstract localization spaces can be considered, such as the location of a data byte in a hierarchical database or a URL on the Internet. In these cases too, there can be some uncertainty in position; for example if position is described by a fuzzy matching rule or an ambiguous data query. In this paper we demonstrate a cooperative transportation task in a group of simulated mobile robots that communi- cate by leaving landmarks in shared localization space. The method is shown to be robust with respect to significant lo- calization error; indicating that it should be suitable for use in real robots.