Learning to Infer the Status of Heavy-Duty Sensors for Energy-Efficient Context-Sensing
ACM Transactions on Intelligent Systems and Technology (TIST)
| Hi-index | 0.00 |
Communicating extensive in-network data generated by resource-constrained wireless sensor nodes is an energy consuming process. To minimise the amount of data exchanged in sensor networks, several researchers have proposed novel and efficient protocols to perform data aggregations, clustering or regression on sensor nodes. Most of these approaches focus on optimising conventional mining techniques to work on resource-constrained sensor nodes. However, the application of association rules for sensor networks is an area of study that has not been investigated. This is due to the high computational cost of obtaining meaningful rules. Thus, in this paper, we propose Adaptive Rule Triggers on Sensors ARTS, to extract highly correlated rules from sensor data and apply them. The learnt rules are used to extend sensor lifetime by controlling sensor operations using triggers. Our approach is optimised to run on non-critical sensing applications/data-aggregation applications that can tolerate a coarse-granularity for sensed data. For this category of applications, our approach can derive meaningful rules efficiently to further conserve energy of wireless sensors. In this paper, these energy savings are evidenced in our experiments that adapt ARTS to a state-of-the-art clustering protocol.