Application of Supervised Learning Approach for Target Localization in Wireless Sensor Network

In the context of indoor environments, the Received Signal Strength Indicator (RSSI) measurements are generally coupled with noise uncertainty due to signal propagation issues such as multipath propagation, Non-Line of Sight (NLOS), reflection. In order to deal with this problem, the localization algorithm is required to be efficient in terms of Localization Accuracy and Execution Speed. The Artificial Neural Network (ANN) does not need prior knowledge of noise statistics during its operations. This paper evaluates the comparison of localization performance of various supervised learning architectures such as Generalized Regression Neural Network (GRNN), Multilayer Perceptron (MLP), Radial Basis Function Network (RBFN), and Feed Forward Neural Network (FFNT) for the Wireless Sensor Network (WSN) based indoor localization problem. The comparison of localization accuracy under the simulated static indoor environment of 100 × 100 m² with 15 anchor nodes advocate the suitability of the application of supervised learning approach for the indoor localization problems over the traditional trilateration-based approach. The proposed supervised learning implementations are tested and compared with the traditional trilateration-based localization technique by varying the variance of RSSI measurement noise from 0 dBm to 5 dBm in the steps of 1 dBm. Out of all the proposed supervised learning architectures, the GRNN based implementation shows higher localization accuracy.