Range feature extraction during active sensor motion

Abstract

An active range sensor is summarised. This sensor can direct its field of view in order to fixate on range features for mobile robot navigation. The image position sensor used has a Gaussian noise characteristic with measurable variance, which makes the sensor particularly amenable to stochastic range feature detection. A geometric analysis of the sensor allows a mathematical model of the sensor to be built, the parameters of which can be determined from data collected during the calibration of the real sensor. This model forms the basis of a sensor simulation, which allows feature extraction algorithms to be developed. One such algorithm, based on the extended Kalman filter, extracts a piecewise-linear range representation of the local environment. This has a number of advantages over previous methods in that it is computationally efficient, it deals with noise appropriately, and it is robust to sensor head movements as range measurements are being made.