Performance evaluation of RSS-based WSN indoor localization scheme using artificial neural network schemes

A popular way of achieving WSN localization is through measurements and evaluation of Received Signal Strength (RSS) values of the signal transmitted by target mobile nodes. However, indoor localization presents a greater challenge due to occurrences of more severe propagation behaviors depending on the parameters of the environment. Artificial Neural Network (ANN) presents a method of adaptive processing of location specific non-linear indoor signal propagation. This paper evaluates the performance of three different methods of ANN family for indoor localization scheme. Data from the simulated propagation model are preprocessed into median, average, min and max values providing a strategic pattern to feed as inputs into the ANNs. The performance of location predicted with Elman Backpropagation (EB), Cascade-Forward Backpropagation (CFB) and Feedforward Backpropagation (FFB) show root mean square error (RMSE) of 0.4991m, 0.5257m and 0.6506m respectively with distance range of 100m.