Clock synchronization is critical to many sensor networks for the success of the application as well as energy efficiency. Achieving a global time frame through localized averaging of clock values for multiple rounds till conver- gence is a promising approach to clock synchronization due to the decentralized nature of computation coupled with scalability. However, it is not clear what power levels for all nodes would make the synchronization process energy- efficient. Large power levels lead to faster convergence but consume a lot of energy per round of synchronization. On the other hand, smaller powers consume little energy per round, but convergence is very slow requiring a lot of rounds to achieve synchronization. In this paper we look at the problem of finding a power assignment that achieves global clock synchronization in the most energy-efficient manner possible. We look at the problem through two dimensions; rate of convergence and energy consumed per round of synchronization. A centralized algorithm is presented that uses the path congestion of the induced communication graph to estimate which power assignments have good convergence properties and find one that minimizes the total energy to achieve clock synchronization. Our evaluation demonstrates that the power assignment derived from this algorithm is very energy-efficient and is applicable for wireless communication environments with various distance-power gradients. Further, we present a simple distributed algorithm which nodes can execute locally to derive energy-efficient power levels for global clock synchronization, and is especially useful in large-scale deployments.
展开▼