Portfolio theory based sensor selection in Wireless Sensor Networks with unreliable observations

In this paper, we propose a portfolio theory based sensor selection framework in Wireless Sensor Networks (WSNs) with unreliable sensor observations for target localization. Fisher information (FI) is used as the sensor selection metric in our work. Our objective is to find a sensor selection scheme that considers both the expected FI gain and the reliability of the sensors, where we observe that the FI variability captures the reliability of the sensors. Based on portfolio theory, we formulate our sensor selection problem as a multiobjective optimization problem (MOP), which is solved by the normal boundary intersection (NBI) method. Simulation results show the advantages of performing portfolio theory based sensor selection.