A Big Data Analysis Framework for Localization in Cloud based Sensor and Robot Networks

Abstract

Internet of Things (IoT) is constructed from billions of sensor devices connected over Internet. Wireless sensor networks (WSNs) are very important communication technologies of IoT for providing large scale data from the environment. Since the data collected grows exponentially, the design of cloud based big data analysis techniques is crucial. Localization in sensor networks is finding the location of a node based on the reference nodes’s coordinates. In many applications such as target tracking and military surveillance, providing localization is necessary. Manually entering the locations of nodes during the deployment phase is not applicable especially for large scale sensor networks. Besides, integrating a GPS receiver to each sensor node is a very costly solution and may not be affordable for large scale networks. Since sensor nodes are mostly battery-powered, design of an energy-efficient localization method is highly desirable to prolong the lifetime of the network. Existing localization techniques may require many message transmission which causes high energy consumption. To tackle with this problem, we propose an energy-efficient localization framework in this paper. A mobile robot is traversed along the sensing area and communicate with sensor nodes to localize these nodes. Different than the previous approaches, our proposed approach requires only 3 messages per node. Besides, most of the execution needed for the localization is not carried by ordinary nodes. We simulate our proposed approach in ns2 simulator. We measure the localization quality and energy consumptions of our proposed approach with its counterparts. Also, we measure the localization quality and energy consumption against varying node counts and degrees. From extensive simulation results, we obtain that the localization qualities of our proposed approach is significantly better than its competitors. Besides, the energy consumption of our proposed algorithm is just 0.06 J per node and far more better than the distributed algorithm. Conclusively, our proposed framework is a significant candidate for IoT and big data applications requiring energy-efficient localization.