Um arcabouço probabilístico para localização dos nós de uma rede de sensores sem fio baseada em RSSI

Abstract

The localization of wireless sensors network (WSN) nodes is a crucial task because in most applications the data they produce needs to be grounded with the position where they were generated. Though there are many ways to locate sensor nodes, there are situations where the use of the received signal strength indicator (RSSI) is the single option. However, there are some difficulties in using the radio location, due to static and dynamic obstruction in the environment. In this work, a probabilistic framework that integrates the different sources of uncertainty is proposed. The methodology developed here addresses the problem of localization of sensor nodes based on RSSI and a precisely located robot. We propose two models that have been tested with different mechanisms of inference. These models establish a relationship with the RSSI and the precise position of a mobile robot through observation models to create a belief state about the location of the sensor nodes. The methodology presented here was evaluated through simulations that allowed controlling the possible variations of factors may be found in real environments. All of these variations are called scenario and is defined by the variation of the noise level in the radio signal, the interval between acquisitions of RSSI samples and the position of sensor nodes. One contribution of this dissertation is the analysis of the interaction between the scenario variation and the sensor nodes localization results, this analysis aims to show which factors affect the results. The experiments showed that the developed framework has other contributions, such as the impact on the evaluation of various scenarios characteristics in the localization process and the good results of locating the sensor nodes with location uncertainty of centimeters in an environment with 7000 m2. Therefore, it can be concluded that the localization of sensor nodes based on RSSI provides a way to locate when the most common technologies are not available, and is able to simultaneously obtain acceptable localization errors in most applications. However, the localization error also increases with increasing disturbance in the signal and the solution consists in the balance of the model parameters to reduce the uncertainty of the process.