Improved techniques for time synchronisation over WiFi and wireless sensor networks

Abstract

Time synchronisation plays a key and ever increasing role in this progressively networked world. Whilst time synchronisation has always been important in industrial and telecommunication systems, its importance has now extended to more diverse applications such as consumer multimedia, medical informatics, SmartGrid, and environmental monitoring. As the need for time synchronisation has grown, so too has the need for such time-sensitive applications to be deployed over wireless networks. It is estimated that there are more than twelve and a half billion Internet-enabled devices connected at the moment - this is expected to reach somewhere between twenty and fifty billion by 2020, as the so-called `Internet-of-things' evolves. Much of this connectivity will be via wireless. The rapid expansion of wireless networks in the last two decades has drastically transformed the architecture of many traditional packet-switched networks. This is particularly the case with TCP/IP based Local Area Networks (LANs) which form a significant building block of the internet. Wireless networks have become ubiquitous within the internet, key amongst them being 802.11 based technologies that now commonly represent the last hop. Whilst hugely convenient, such networks present many new challenges in terms of time synchronisation. In particular, 802.11 or WiFi marks a return to contention based access which can lead to significant delays that are often asymmetric. This greatly degrades the performance of common synchronisation protocols that are designed for largely symmetric wired networks. Another wireless based network has experienced significant, albeit less spectacular growth in the last decade, namely, Wireless Sensor Net- works (WSNs). They provide a means to gather physical data via a wireless network of low-powered, inexpensive computing devices. These miniature devices, although limited in terms of power and computing resources, have huge potential when used in a distributed fashion. They can be used to collect sensory data from large geographical areas which, when collated, can provide invaluable insight into a phenomenon under observation. Data collected via WSNs often requires precise time stamping, and thus, in such cases, WSNs require time synchronisation protocols in order to operate effectively. Since the computing devices that compose a WSN are typically power-limited, the overhead of a time synchronisation protocol can use up valuable energy, thus, limiting node life. The research contributions in this thesis address key synchronisation problems within both 802.11 and WSN domains. The significant problem concerning the degradation of time protocols operating over 802.11 networks is remedied via a novel mechanism that exploits the standard operation of 802.11 networks in order to reduce the error in datasets employed by such protocols. Validation experiments performed confirm that the mechanism can deliver synchronisation accuracies akin to those achievable over wired networks. In relation to WSNs, the research contribution addresses the key problem of communication (and thus energy) overhead for synchronisation protocols. This contribution is presented by means of a new protocol, termed the Dynamic Flooding Time Synchronisation Protocol (D- FTSP) which offers an energy efficient means of synchronising WSNs deployed in dynamic environments. Experiments reveal that D-FTSP, in particular scenarios, can result in significant energy savings with respect to alternative time protocols, thus, greatly extending the life of a WSN.