Data fusion in 3D through surface tracking

Sensory signals typically suffer from noise, ambiguity, spurious signals and omissions. For a robot to successfully model the environment in which it operates, it must use signals captured from different locations in time and space in an effort to select those signals that appear to be accurate. Data fusion, the process of combining signals into a single representation, is an essential component of a mobile robotics system. In this paper, we describe a new method for solving data fusion for a typical mobile robot domain, one in which precise robot location information in not known and where the robot mounted sensors employed are not calibrated.