Otimização do método de localização de Monte Carlo para redes de sensores sem fio móveis
Matos, João Otávio Cadó de
The increasing development of electronic circuits integration technology combined with the evolution of the microelectromechanical systems has enabled rapid advancement of research in the area of wireless sensor networks, which are composed by a set of sensor de-vices with extremely limited resources in the processing power, data storage capacity and energy consumption. Thanks to these advances, wireless sensor networks are being used in an increasing number of applications involving military and medical activities, environmental, domestic and industrial monitoring, traffic control, among others. In many of these applica-tions, the knowledge of the location of the sensor nodes is a fundamental require, besides con-tribute to increasing the safety and lifetime of the network. The use of GPS devices to meet such a localization function, however, becomes impractical as the network nodes density in-creases, in view of the increasing cost of implementation and power consumption. Thus, sev-eral algorithms are proposed to cheapen and/or improve the precision of the nodes localization process in wireless sensor networks, whether mobile or static. Therefore, this work presents a node localization scheme based on Monte Carlo localization algorithm for mobile wireless sensor networks. With the proposed amendments on such an algorithm, the main objective was to reduce the estimate location error, especially in scenarios with low GPS devices densi-ty. To evaluate the behavior of the proposed scheme estimate error against some parameters of the network and the method itself, it was developed a simulation tool in MatLab software, which has proved quite versatile, given the parameters to manipulate and analyze, and easy to use. Thus, the proposed localization scheme was simulated and compared to some of the main localization algorithms derived from Monte Carlo method for wireless sensor networks, adopting the estimate location error as the comparison metric. Considering scenarios with different GPS devices densities and with nodes moving according to a modified version of the random waypoint mobility model, the proposed localization scheme presented a location error lower than that observed for the other methods analyzed.