Algorithmic Methodologies for De-Centralising Time

About the Project

This project represents a development in a truly sparse capability within the UK. With the current, very rapid development of requirements for future timing systems, in both the military and civil arenas, there is a keen and growing demand for suitably qualified and experienced persons in these areas to which the work in this proposal will contribute significantly. Only one studentship is available, recruitment will close on selection of a suitable student.

Synchronising clocks represents a challenge which until recently, was typically solved through centralised means. This project will explore ensemble time; the use of Wireless Sensor Networks, WSNs, and Internet of Things, IoTs, to fuse highly accurate atomic clocks with lower quality clocks, e.g., quartz. Enabling both large and small groupings of time synchronised nodes to maintain highly robust time synchronisation, via methods of decentralisation that are suitable for complex, time-varying, and challenged network topologies.

This project will develop algorithmic approaches to ensemble time, and its full decentralisation. Decentralised synchronisation depends on the correction of clock drift, from a given reference time or rate of time. Distributed consensus mechanisms operating over a communication network can correct for drift; a product of phase noise, jitter, aging, and manufacturing variance. Here, network nodes share and compare their local time in order to agree on a shared consensus time. Each node maintains their own virtualised time opinion that is steered towards the consensus time, by varying an offset and ticking rate with reference to their immutable hardware time. The integration and application of resilient consensus algorithms establishes a decentralised timing network.

As WSNs and IoTs have become increasingly prevalent, so have Distributed Network Time Synchronisation, DNTS, methods and their associated consensus protocols. This project will develop algorithms for time decentralisation in dynamic communication topologies, where network nodes move with respect to each other; disconnecting and reconnecting to create time-varying, ad-hoc mesh networks. The project will consider time consensus and synchronisation in terrestrial and space networks.

DNTS methods attempt to bring network nodes to a synchronised time without the need for a central authority or parent-child commands, offering an alternative to GNSS methods. Consensus methods are a typical approach for such time synchronisation, although resilience is a challenge that remains largely unaddressed by such algorithms, and is a key aspect of this project. The network must cope with, and/or recover from disruptions to be robust and secure. Such malfunctions may come from system faults or security threats. For synchronisation to be fault tolerant, it must be robust against a range of fault manifestations, dominant among these will be transient communication failures. Due to the importance of this work there will be significant and energetic support from the industry partner, including support with timing transfer experimentation and opportunities to test algorithmic capability on experimental systems, and even on mobile assets.

Funding Notes

This is an EPSRC Industrial Doctoral Landscape Award (IDLA), sponsored by Dstl, which covers the 4-year value of stipend, fees, travel and equipment costs. A minimum, tax-free, stipend of £19,500 per year is available, and an enhanced, tax-free, stipend is available to qualifying candidates. A travel and training budget of, at least, £15,000 will be provided to the candidate.