Applied Sensor-Assisted Monte Carlo Localization for Mobile Wireless Sensor Networks

Localization is a mandatory requirement in almost all sensor network applications. In order to avoid power-hungry GNSS solutions like GPS, alternative localization algorithms based on anchor nodes are developed. Unfortunately, many proposed solutions focus on static networks and do not account for mobility. An often cited and further improved approach is Monte Carlo Localization (MCL), which is one of the first methods accounting for total mobility in the network. In this paper we propose and practically evaluate Sensor-Assisted Monte Carlo Localization (SA-MCL), which is designed to bypass situations of temporary connection loss to anchor nodes due to changing network topologies. SA-MCL uses additional sensor information to update the position estimation of a node in case no anchor nodes are available. We show by both practical field tests and simulations that our dead reckoning approach can successfully account for situations without anchor node information and reduces the localization error by about 58 %. Our practical implementation is performed on IRIS sensor motes which are mounted to the backside of radio controlled cars.