RSSI-based node localization in wireless sensor networks using a hybrid approach

Abstract

Localization of nodes is an important issue in Wireless Sensor Networks (WSNs), that has attracted the interest of the scientific community. Localization comprises the process of determining the position of a sensor node in a WSN. A WSN is made up of a large number of tiny, low-energy, limited-processing-capability, low-cost sensors that communicate with one another and form an ad-hoc network capable of getting across relevant data and messages. Accurate location of nodes in a WSN is non-trivial given that WSNs are often deployed in locations not easily accessible by GPS and that GPS sensors are heavy and dicult to deploy on such resource constrained nodes. This thesis focuses on a hybrid localization scheme for WSN nodes that makes use of a machine learning algorithms (Random Forest in this case) and multilateration. The position of a sensor node is computed by combining both algorithms. The approach involves using a pre-trained machine learning algorithm to locate a large number of unknown nodes that fall within the communication range of pre-localised nodes called anchor nodes. Subsequently, these newly localized nodes serve as newly added anchor nodes and a simple multilateration algorithm is used to localise subsequent nodes. The multilateration algorithm is harnessed for several iterations until all the nodes in the region of interest are localised. The ecacy of the proposed approach for localization in WSN is validated first using simulated datasets. The results show that the proposed approach outperforms a variety of other approaches to localizing the WSN, including Support Vector Regression, Decision Tree, Bossting Techniques, and Neural Networks. Following the validation using simulated datasets, we demonstrate the approach’s ecacy in a real-world environment comprising a prototypical WSN deployment. The approach is found to work well and the localization in the real world largely matches that in the simulated environment. Keywords: Localization, Wireless Sensor Network, Range-based localization, RSSI, Random Forest, Iterative Multilateration