Thesis Summary

 

CLOCK SYNCHRONIZATION

IN WIRELESS SENSOR NETWORKS

 

Kasım Sinan YILDIRIM

 

PhD. in Computer Engineering

Supervisor: Assoc. Prof. Dr. Aylin KANTARCI

12.04.2012

 

 

Sensor nodes in wireless sensor networks (WSNs) are equipped with cheap hardware clocks which frequently drift apart due to their low-end quartz crystals whose frequencies are not stable and vary over time. Environmental factors such as temperature changes result in short-term frequency instability and subtle effects such as oscillator aging result in long-term frequency instability. Hence, the hardware clocks of the nodes may not remain always synchronized although they might have been synchronized when they are started up.

 

Lack of synchronized time leads to inaccurate and inefficient operation of many applications and protocols in WSNs. For instance, sensor nodes need to access common time for time coordinated actions such as synchronous power on and shutdown of the communication circuit, which reduces energy consumption of the battery-powered sensor nodes. Therefore, a time synchronization protocol is required so that all nodes exchange their time information to synchronize their clocks for minimizing their synchronization error, i.e. clock skew.

 

A common strategy in order to achieve network-wide time synchronization is to flood current time information of a reference node into the network, which is utilized by the de facto time synchronization protocol Flooding Time Synchronization Protocol (FTSP). If synchronization to stable time sources such as Coordinated Universal Time (UTC) is required, employing the method of flooding in order to provide time synchronization also becomes crucial. In my PhD thesis, we firstly considered flooding based time synchronization and revealed the drawbacks of the existing synchronization schemes which are based on flooding. Within this context, my thesis presents the following two contributions for flooding based time synchronization in WSNs:

 

 

While studying flooding based time synchronization protocols, we revealed that the least-squares mechanism employed by these protocols may set the clocks of the nodes back. Hence, my thesis revealed this important problem as a contribution to the time synchronization research:

 

 

 

Flooding based synchronization schemes may poorly  synchronize neighboring nodes. Hence, another perspective in time synchronization in WSNs is to provide tight synchronization among the neighboring nodes, namely gradient clock synchronization. However, we revealed that there is a lack in the literature of a gradient time synchronization protocol  which can provide external clock synchronization. Within this context, my thesis presented the first external gradient time synchronization protocol in the WSN literature:

 

 

Apart from the practical side of the gradient clock synchronization, there are several theoretical studies which propose theoretical gradient clock synchronization algorithms. We  studied these algorithms and revealed that the CPU overhead of the optimal gradient clock synchronization algorithms in the literature can be reduced. As a last contribution of my thesis, we presented  a new optimal gradient clock synchronization algorithm which is more efficient in terms of CPU consumption: