Simultaneous Localization and Mapping (SLAM)

SLAM is a very active area in the robotics research community. There are many approaches to it using a great variety of sensors and algorithms. We will focus on the case where only cameras are used, and we will explore how the Kinect simplifies the SLAM problem.

Imagine a robot that is being dragged across a room and whose goal is to return to its initial position afterward. As it is being moved across the room, it has only cameras to learn the path back home. This poses a great challenge, because it must map the structure of the room and determine its changing position simultaneously. As we have seen before, we can obtain feature matches between images as the robot changes position, but the 2-D registration techniques we studied cannot map its 3-D motion across the room.