Camera-aided human navigation: Advances and challenges

We examine the use of visual navigation techniques for improving human navigation and localization in GPS-denied environments. While many advances have been made in navigation techniques using stereo and monocular motion estimation, significant hurdles exist for a practical system, including size, weight, power, and cost limitations. Humans often move and rotate faster and with more complex motions than robots, therefore requiring increased processing speed and robustness and the use of specialized algorithms. Systems for long-range navigation in previously unmapped environments must deal with error drift. Monocular systems face the additional issue of scale drift, since the relative scale must be estimated repeatedly. Furthermore, the orientation of the system relative to the frame-of-reference must be initialized accurately. We have evaluated these issues using a simulator and discuss possible solutions.