Node Localization in Wireless Sensor Networks Using Multi-output Random Forest Regression

More advanced developments have been made in the field of wireless communications, and it has further accelerated the growth of compact and low power-consuming wireless sensor nodes. During communication, each source node estimates the shortest path to the destination node by using the location information. Location information also helps in securing the network in the prevention of intruders. Previously available sensor node localization methods in the literature such as radio signals, time of arrival (ToA), and time difference of arrival (TDoA) suffers from various drawbacks. Also, the usage of sophisticated devices like GPS to sense the location of the node increases the deployment cost and in parallel, the energy consumption is also increased. This paper aims at developing a model to predict the future location of a dynamic sensor node. The linear model is built using the historical location information of the respective node. The trained model is capable of predicting the X- and Y-coordinates of a node accurately. For each of the node, a separate model is built and their future locations are predicted. If a node has data packets to transmit to a sink node, it obtains the present and next location of the sink node from the base node.