Energy-Aware Measurement Scheduling in WSNs Used in AAL Applications

In wireless sensor networks developed for ambient assisted living applications, the supply of the required power is one of the most challenging problems. Batteries have remarkable drawbacks, and in some cases, the change of batteries is impossible (space, infected area, etc.). We approached the problem from two directions: 1) The energy for the sensor node's operation should be harvested from the environment, and 2) the nodes should work as efficiently as possible. A new method is presented, which optimizes the whole network energy demand while maintaining the performance of the system, with the scheduling of the measurements and sensors. It is taken into account that both the measurement of a physical variable and the transmission of a message have different costs. Selection of sensors and measurement intervals in the system is based on a cost assigned to each sensor, which considers 1) the estimated state of the observed variable based on the past measurements and a model, 2) the actual energy state of the sensor, and 3) the possible future events that will affect the energy levels and/or the observed variable. A hidden Markov model is used to assign probabilities to the states of the unknown variables, which are to be observed. The probabilities of the state transitions are specified by a learning process. Then, a defined cost function is applied to calculate the cost of each sensor, the sensors with the minimal cost will be configured for more frequent measurements ensuring precision, and the others will be configured to less frequent measurements to save energy.