The dawn of large language models like BERT and GPT-3 is set to unleash a revolution in the way we experience life. Working, learning, and interacting with humans and the digital world will never be the same again as these models pave the way to a new era. As we entrust these powerful models with critical tasks and sensitive decision-making, ensuring their unwavering performance becomes an absolute
necessity. However, these systems suffer from non-robust performance, casting a shadow of
uncertainty over their potential.
This project aims to fortify the robustness of large language models and certify their performance.
Intelligent agriculture technology has the potential to revolutionize today's agriculture practices and has been held as a crucial component of sustainable agriculture. iAgriculture DMS aims to develop the necessary data management technology to support user-friendly and effective intelligent agriculture technology with an emphasis on precision irrigation.
This interdisciplinary project focuses on controlling the charging of plug-in electric vehicles (PEVs) in residential and small commercial settings using a novel and flexible open-source, open-architecture charge communication, and control platform. This software-based platform will be embedded in the context of overall utility and residential/business electrical and building automation systems, lending itself to potential broad implementation by commercial interests. The project also focuses on the key issues associated with the development of the open-source platform including assessment of user needs and grid operation and ratepayer benefits, grid security considerations, and the potential for PEV charge control to lead to increased ability to accept intermittent renewable energy for California’s electrical grid.
An award-winning, EPSRC-supported, multi-million pound, collaborative research endeavor with the mission of delivering the science of human-agent collectives and applying it to the critical domains of the smart grid, disaster response, and citizen science. In this context, ORCHID explored the features of flexible autonomy, agile teaming, incentive engineering, accountable information infrastructure. A key component of ORCHID was the close connection and inter-play between world-leading fundamental research, the demonstration of this research in compelling real-world application scenarios, and the involvement of collaborating partners whose future lies in exploiting such research in these application areas.
Recycling cardboard and paper has been heralded as a key tool towards
a sustainable future. However, it involves a number of labor-intensive
steps. Once recyclable materials reach the recycling facilities, cardboard
and paper must be separated from the rest of the recyclable material and
contamination needs to be identified. Contaminated cardboard cannot be
recycled, and can even ruin clean batches of recycled cardboard. This project proposes a full intelligent pipeline for
recognizing cardboard and paper, as well as identifying contamination based on machine vision techniques.
Precision irrigation scheduling is a vital farm water conservation practice and considers a challenging decision-making problem that determines when and how much to irrigate. The process is copious and error-prone and its automation considers a valuable link toward sustainable water conservation practices. This project aims to develop an intelligent method and a corresponding online tool that supports such decision-making in pistachio plants. To this end, it develops image recognition and machine learning techniques to determine the irrigation plant needs by identifying the respective canopy temperature. We anticipate the tool to be used by the scientific community and the wider public.
A CEC-supported, collaborative, multi-million dollar research project with the mission to enable intelligent heating, ventilation and air conditioning control in low-income households through the development of an appropriate low-cost smart thermostat.
Estimating the power output of inherently intermittent and potentially distributed renewable energy sources is a major scientific and societal concern. RENES (Renewable Energy Estimator) is an award-winning interactive web-based tool that enables short-to-mid term forecasts of photovoltaic (PV) systems and wind turbine generators output. Renes comes complete with a web-based application program interface (API), thus enabling the user to fully exploit the tool's requests. In this context, RENES is employed by Artificial Intelligence and Multiagent Systems researchers, for activities such as accumulating data, designing experiments, or evaluating new algorithms. RENES is not a commercial product and is available to all, free of charge. Its traffic picked at ∼10, 000 clicks/day before respective constraints were put in place at the end of 2016 to keep the service available free-of-charge
Small farmers today do not employ automated irrigation systems as the cost is typically prohibitive to them. This project aims at promoting irrigation automation to conserve on-farm resources via the development of a low-cost, expandable, open-source farm network infrastructure and operating system to primarily support efficient irrigation practices.
The services demanded of commercial building customers—heating, cooling, ventilating, lighting, computing, and plug loads—require significant energy and contribute to peak energy demand. The project goal is to improve energy efficiency by enabling effective management and integration of demand response associated with tariff schedules and distributed generation with other building services in residential and commercial buildings by developing a system architecture supporting demand-response message passing and translation between the smart grid and the XBOS-DR building management system.
This interdisciplinary project aimed at developing machine learning-based gravitational wave identification methods to empower the LIGO pipeline. It succeeded a sensitivity study on r-mode gravitational wave signals from newborn neutron stars illustrating the applicability of machine learning algorithms for the detection of the long-lived gravitational-wave transient.