WiFi position estimation in industrial environments using Gaussian processes

The increased popularity of wireless networks has enabled the development of localization techniques that rely on WiFi signal strength. These systems are cheap, effective, and require no modifications to the environment. In this paper, we present a WiFi localization algorithm that generates WiFi maps using Gaussian process regression, and then estimates the global position of an autonomous vehicle in an industrial environment using a particle filter. This estimate can be used for bootstrapping a higher-resolution localizer, or for cross-checking and localization redundancy. The system has been designed to operate both indoors and outdoors, using only the existing wireless infrastructure. It has been integrated with an existing laser-beacon localizer to aid during initialization and for recovery after a failure. Experiments conducted at an industrial site using a large forklift-type autonomous vehicle are presented.