Artificial intelligence design aided photonic systems

Integrated photonics has proven to be the ideal solution for tackling old scientific challenges and exploring new physics and technologies. The increasing demand for performance and circuits complexity has added pressure to the computational effort required for the design and the fabrication tolerances. Since Machine Learning (ML) has been used to generate novel designs, there has been a race to utilise such a tool as a fast, efficient, to some extent, superior design entity. The research objective of this project is to develop trained ML models capable of providing the proper integrated photonics systems structure/configuration starting from the desired optical performance specification. The inverse design problem, meaning the definition of device features starting from the desired functional parameters, will be approached using ML in combination with simulated populations of devices in the first instance. This method will enable the modelling of non-intuitive designs. The designs on the chip will include integrated photonic circuits with different levels of complexity: spanning from passive systems (splitters, WDM, etc.) to integrated programmable photonic networks. The approach and methodology for the project consist of the use of ML as an efficient method to design photonic devices and systems. The projects will rely on data collection from simulation (in the first instance) to develop data-driven design strategies instead of the physics rule models methods. The data collection will be fed to ML algorithms and software agents, which, after proper training, will then provide features for passive devices and electrical driving sets for programmable systems. The project will target devices with (partially) known physical rules (splitters, wavelength filters) to assess and benchmark the data-driven models. The project will then target structures for which the traditional approach fails or is not reasonably feasible: an example of this activity is the design of filter/splitters and the driving of programmable photonic circuits. The most ambitious and supreme goal is to use the ML and artificial neural network (ANN) models to design novel Optical Neural Networks (ONN).

The impact of the CDT in Connected Electronic and Photonic Systems is expected to be wide ranging and include both scientific research and industry outcomes. In terms of academia, it is envisaged that there will be a growing range of research activity in this converged field in coming years, and so the research students should not only have opportunities to continue their work as research fellows, but also to increasingly find posts as academics and indeed in policy advice and consulting.

The main area of impact, however, is expected to be industrial manufacturing and service industries. Relevant industries will include those involved in all areas of Information and Communication Technologies (ICT), together with printing, consumer electronics, construction, infrastructure, defence, energy, engineering, security, medicine and indeed systems companies providing information systems, for example for the financial, retail and medical sectors. Such industries will be at the heart of the digital economy, energy, healthcare, security and manufacturing fields. These industries have huge markets, for example the global consumer electronics market is expected to reach $2.97 trillion in 2020. The photonics sector itself represents a huge enterprise. The global photonics market was $510B in 2013 and is expected to grow to $766 billion in 2020. The UK has the fifth largest manufacturing base in electronics in the world, with annual turnover of £78 billion and employing 800,000 people (TechUK 2016). The UK photonics industry is also world leading with annual turnover of over £10.5 billion, employing 70,000 people and showing sustained growth of 6% to 8% per year over the last three decades (Hansard, 25 January 2017 Col. 122WH). As well as involving large companies, such as Airbus, Leonardo and ARM, there are over 10,000 UK SMEs in the electronics and photonics manufacturing sector, according to Innovate UK. Evidence of the entrepreneurial culture that exists and the potential for benefit to the UK economy from establishing the CDT includes the founding of companies such as Smart Holograms, PervasID, Light Blue Optics, Zinwave, Eight19 and Photon Design by staff and our former PhD students. Indeed, over 20 companies have been spun out in the last 10 years from the groups proposing this CDT.

The success of these industries has depended upon the availability of highly skilled researchers to drive innovation and competitive edge. 70% of survey respondents in the Hennik Annual Manufacturing Report 2017 reported difficulty in recruiting suitably skilled workers. Contributing to meeting this acute need will be the primary impact of the CEPS CDT.

Centre research activities will contribute very strongly to research impact in the ICT area (Internet of Things (IoT), data centre interconnects, next generation access technologies, 5G+ network backhaul, converged photonic/electronic integration, quantum information processing etc), underpinning the Information and Communications Technologies (ICT) and Digital Economy themes and contributing strongly to the themes of Energy (low energy lighting, low energy large area photonic/electronics for e-posters and window shading, photovoltaics, energy efficient displays), Manufacturing the Future (integrated photonic and electronic circuits, smart materials processing with photonics, embedded intelligence and interconnects for Industry 4.0), Quantum Technologies (device and systems integration for quantum communications and information processing) Healthcare Technologies (optical coherence tomography, discrete and real time biosensing, personalised healthcare), Global Uncertainties and Living with Environmental Change (resilient converged communications, advanced sensing systems incorporating electronics with photonics).