Location agent: a study using different wireless protocols for indoor localization

Context-aware systems have received greater interest in the computing community. In order to provide relevant services at context-aware applications, the first task is to locate the user, what can be done preferably dynamically and intelligently. However, indoor mobile users localization is not a trivial problem, since it involves checking various devices, transmitting signals simultaneously on the same radio frequency, with possibly the three existing wireless network protocols: Wi-Fi, Bluetooth and ZigBee. In this direction, this paper presents an agent-based architecture with the Location Agent module defined for context-aware applications that uses three artificial neural network algorithms trained for the different protocols: backpropagation, backpropagation with momentum and levenberg–marquardt. Considering the research experimental aspects, a study is presented to compare the neural network algorithms including performance, regression analysis, precision and accuracy. The results indicate that the backpropagation algorithm trained with Bluetooth provides better accuracy (the average error of 0.42 meters) and the backpropagation trained with Wi-Fi provides better precision (73%). We consider our approach promising since the Location Agent has a quality of service component associated with the neural network algorithms that can choose the best received signal strength to locate indoor users.