The Reconfiguration Problem in Sensor Networks: An Optimization Approach

The objective of reconfiguration in sensor networks is to maximize the overall network lifetime while maintaining a minimum quality of service. This is performed by turning some of the nodes off whenever they are not needed to maintain the required quality of service. This requires a way of detecting node failures to reconfigure accordingly. The reconfiguration problem can be formulated as an integer linear programming optimization problem. A solution to this problem consists of a binary vector, where a "0" in position k means that the corresponding node can turn itself off in the next reconfiguration period, while a "1" means that the node is required to remain active. A centralized solution to this problem is designed and implemented. However, this solution suffers high energy costs and long delays. In addition, it does not scale well with the network size. A cluster-based solution is derived, and we are currently working on a localized solution where nodes interact only with neighbors to decide whether to turn off or remain active.