| This work presents a new methodology for controlling a team of small-scale, resource-limited robots in an autonomous exploration task. Small robots by their very nature are limited in their perspective, the quality of information they can collect and the ability to move around. However, by collaborating both sensing and movement, a team of small robots can extend sensor range and resolution beyond that of any individual to move competently through a cluttered field of obstacles.; Managing a team of small robots naturally inherits all the difficulties of the individual---specifically the limited ability to gather data. Research with small robots is made even more challenging by the fact that teams critically depend on each other while at other times interfere with one another. In this thesis, we examine the challenges of developing control algorithms for multiple, small-scale, resource-limited robots that coordinate motion and share sensor information.; We start by exposing the inherent weaknesses of conventional free space approaches to sensor error and uncertainty. We more closely examine the way sensors produce errors for small robots and how these errors impact the way the team builds maps and decides how to move. We base this work on the underlying fact that sensors detect obstacles not free space and utilize this insight to propose alternative methodologies that are robust in the face of increasing sensor uncertainty. While this work can improve the effectiveness of an individual robot, it becomes especially pertinent as robots become resource-limited.; We also propose methods for augmenting the information stored in the team map. Exploiting information implicitly collected in maps allows the team to proactively position itself in ways that both improve localization and increase the resolution of existing knowledge about a space. These collaboration methods more closely leverage the geometry of the team to plan movement in such a way as to maximize information gain while minimizing localization error. By combining these approaches, we can achieve autonomous exploration with a team of small-scale, resource-limited robots. |
