A POMDP framework for data acquisition in wireless sensor networks

This paper proposes a sensor selection scheme for data acquisition which supports probabilistic confidence requirements in wireless sensor networks (WSNs). The aim of the scheme is to optimize the long-term performance criterion in data collection while maintaining data reliability under error-prone WSNs. We formulate the problem as a partially observable Markov decision process (POMDP). An existing tool used for solving POMDP called the witness algorithm is then employed to find an optimal sensor selection policy solution such that the requirements of the quality assurance on the sensor readings are still satisfied. Simulation results show that the proposed scheme can achieve the optimal long-term performance criterion as well as save energy consumption when compared to previously proposed data acquisition methods.