Improved SOM-based distributed range-free localization for mobile sensor networks

Recent researches in mobile wireless ad-hoc and mobile sensor networks have shown the importance of location information from nodes participating in the networks. Although nodes with GPS-equipped device (anchor nodes) can be deployed in those networks to obtain the location information from nodes, researchers are interested in solving localization in networks that have a small fraction of anchor nodes deployed due to the high cost and energy expenditure of GPS devices and indoor environments. Most of existing researches focus on static networks, and only a few researches focus on mobility problem of practical scenarios. Localization schemes dealing with mobility problem usually apply the sequential Monte Carlo algorithm to solve the mobility problem and they put high computational load on network nodes. This paper proposes a solution to improve our existing method that is based on Self-Organizing Maps (SOM) to make it adaptive to high mobility of nodes. The proposed solution improves the location estimation of nodes by combining location estimation based on previous observations with new information observed in the current step during SOM learning. Simulation results from both purpose-built program and network simulator 2 (NS-2) confirm the effectiveness of the proposed solution.