As digital technology permeates every area of modern life, we risk becoming over-dependent on complex systems that operate in an opaque way, creating a risk that they exhibit emergent properties that adversely affect their users or their wider environment. This is particularly true as developers increasingly rely on AI or ML techniques as a means to define system behaviour when the problem space is too complex or poorly understood for human developers to explicitly specify that behaviour. We are tackling incompletely understood problems by developing systems whose behaviour and wider impact are by necessity also incompletely understood. This trend, which is largely enabled by an abundance of data harvested from (e.g.) mobile devices, sensors and social media, is radically changing how systems are developed and how they are used. We need a new approach to software engineering that (i) places greater emphasis on making explicit the risks of unintended behaviour for innovative new software products either through limitations on our understanding of the envisioned product's behaviour or through misuse, and (ii) actively supports explainability of the exposed behaviour by the running system. Twenty20Insight is an interdisciplinary project bringing together academic experts in Software Engineering (SE), RE, Design Thinking and ML to help system stakeholders and developers understand and reason about the impact of intelligent systems on the world in which they operate.

Unlike previous generations of mobile networks, the beyond 5G (B5G) network is envisioned to support edge intelligence, which is to provide both communication and computing capabilities to the proximity of end users. Wireless edge intelligence is particularly important to those crucial use cases of B5G, including smart cities, autonomous driving, wireless healthcare, virtual reality (VR) and augmented reality (AR) gaming, where mobile networks are expected to be equipped with intelligent capabilities for prediction and shaping experiences to individuals. Federated learning (FL) is a key enabling technology for wireless edge intelligence, by performing the model training in a decentralized manner and keeping the data where it is generated. However, a straightforward adaption of FL from computer networks to wireless systems can suffer performance degradation in spectral and implementation efficiency, because of the complex wireless environment with heterogeneous resources and a massive number of devices. The aim of this project is to develop a novel scalable hybrid architecture for wireless FL by efficiently utilising the physical layer dynamics of the mobile communication environments and exploiting sophisticated service-aware and resource-aware collaborative edge learning. The novelty of the project is the development of this novel edge learning architecture, where the fundamental limits of the learning architecture is characterised by advanced mathematical tools, such as graph theory and stochastic learning. In addition, an algorithmic framework for quantifying challenging design trade-offs in the presence of practical constraints by applying sophisticated tools such as compressed sensing and machine learning.

The aim of this proposed research is to address the modelling, design, demonstration and potential applications of ultra-wide-band (UWB) optical fibre amplifiers based on the Raman effect, induced by high power laser pumping of specially designed optical fibre, for future applications in optical fibre communication networks, ranging from inter-data-centre connections to metro/regional networks. Despite massive advances in the capabilities of optical fibre communication systems over the past two decades, enabled by digital coherent technology, internet traffic growth remains well above 20% per annum, and is forecast to continue on a strong trajectory for the foreseeable future. Delivering a seamless optical amplifier of unprecedented bandwidth is now seen by operators and their network equipment suppliers as the most practical and cost-effective way to increase the traffic carrying capacity of the billions of km of glass fibre that has been deployed worldwide, by making use of the wide low-loss window. The programme targets two specific designs of all-Raman amplifier: (i) a node-located, discrete-only parallel, dual-stage design, and (ii) a hybrid distributed-discrete dual-stage design, making use of the intra-node transmission fibre as a gain medium for part of the spectrum. These innovative designs are enabled by recent increases in laser pump powers and novel nonlinear Raman gain fibres, and a growing, general acceptance of Raman technology by all network operators, ranging from relatively conservative incumbents, such as Verizon, to more adventurous technology giants, such as google. New, nonlinear, modelling tools will be developed to overcome and support the significant experimental design challenges in manufacturing and operating our proposed UWB amplifiers, which with 300nm bandwidth offer approaching 10x the bandwidth of standard Erbium-doped fibre amplifiers used in today's networks. Key optical amplifier characteristics such as gain, noise figure, uniformity and nonlinearity will be measured stand-alone. UWB optical fibre communication system capacity improvements and performance will be evaluated in representative models of target networks, informed by our project partners, and compared with extensive in-line and recirculating loop UWB laboratory-based tests.

The UK, Ireland, Canada and France have all declared climate emergencies. Climate change has never had a more prominent in the public eye. With legal commitments to reduce greenhouse gas emissions by at least 80% by 2050 relative to 1990 levels, it has never been more important to do everything we can to reduce energy demand. The promise in this project is to help provide new methods for analysing the 'data deluge' of energy and building system data (from IoT devices) that can help unlock energy efficiencies and identify the benefits of energy efficiency measures despite noisy and heterogeneous data; and make it cheap, repeatable and routine to do this on an ongoing basis. Key to our approach are novel statistical and mixed-method techniques working closely with our project partners and their data to demonstrate the feasibility of these benefits. Our ultimate goal is to make it possible to translate the savings found in one context to another (e.g. another similar building, or even similar business). This would enable the 'digital replication' of energy efficiency savings, and even an almost viral spread of the knowledge and technique across sectors---with massive potential. Currently for many organisations, making sense of this rich source of information defies the human resource available to analyse and profit from the potential insights available. Such analysis is currently the domain of specialist consultancy providers due to the significant cost, time and know-how required to identify opportunities in the data. This restricts the penetration of data-driven monitoring and energy reduction strategies, and the opportunities for knowledge transfer across different locations and businesses. This project will clear this analysis bottleneck. The approach builds on foundations in modern data science, applying cutting edge techniques to automatically identify problems at particular sites and recommend interventions based on cross-site comparisons. The principle objective is to enable commercial sites to reduce their energy demand and keep it low without requiring energy analysts to manually investigate each site individually, at further expense. Core to our approach are next-generation statistics and machine learning methods applied to a unique corpus of fine-grained energy and process data sourced from our partners (BT, Tesco, Lancaster University Facilities (a town sized campus), and energy management consultancy and cloud energy analytics provider, BEST). This will enable us to apply cutting edge statistical techniques to a very significant data set in this domain for the first time. More specifically, our main aims are to: 1. develop automated techniques for supporting analysis, identifying and recommending energy savings strategies, based on the application of statistical and machine learning techniques to fine-grained energy data; 2. derive knowledge of how, where and when energy is used, to identify opportunities to reduce and shift demand by comparing differences in energy use over time within and between premises; 3. support regular and repeated analysis, towards a continual improvement in energy reduction over time. 4. provide open source, permissively licensed implementations for enabling uptake, even beyound our project partners and their partner networks. Our publication and publicity strategies will maximise exposure of our project results to various stakeholder groups including academia, practitioners, and key industry stakeholders.

As digital technology permeates every area of modern life, we risk becoming over-dependent on complex systems that operate in an opaque way, creating a risk that they exhibit emergent properties that adversely affect their users or their wider environment. This is particularly true as developers increasingly rely on AI or ML techniques as a means to define system behaviour when the problem space is too complex or poorly understood for human developers to explicitly specify that behaviour. We are tackling incompletely understood problems by developing systems whose behaviour and wider impact are by necessity also incompletely understood. This trend, which is largely enabled by an abundance of data harvested from (e.g.) mobile devices, sensors and social media, is radically changing how systems are developed and how they are used. We need a new approach to software engineering that (i) places greater emphasis on making explicit the risks of unintended behaviour for innovative new software products either through limitations on our understanding of the envisioned product's behaviour or through misuse, and (ii) actively supports explainability of the exposed behaviour by the running system. Twenty20Insight is an interdisciplinary project bringing together academic experts in Software Engineering (SE), RE, Design Thinking and ML to help system stakeholders and developers understand and reason about the impact of intelligent systems on the world in which they operate.