Localization with RSSI Values for Wireless Sensor Networks: An Artificial Neural Network Approach

Collecting the data and forwarding it to its destination is an important function of a sensor network. For many applications, it is also very important to include the location information to the collected data. Localization techniques in wireless sensor networks (WSNs), which is a way of determining the sensor node locations can be utilized for obtaining such information. In this paper, we present an artificial feed-forward neural network based approach for node localization using the RSSI values of the anchor node beacons. The method was applied in a 5 x 4 m indoor environment with beacons installed at each of the corners. Five different training algorithms have been evaluated to obtain the training algorithm that gives the best results for this scenario. The multi-layer Perceptron (MLP) neural network has been obtained using the Matlab software and implemented using the Arduino programming language on the mobile node to evaluate its performance in real time environment. An average 2D localization error of 30 cm is obtained using a 12-12-2 feed-forward neural network structure. The method proposed can be used to implement a trained neural network using other programming languages on other platforms.