Approximate Computing for Power and Energy Optimisation

Informazioni relative al progetto

APROPOS

ID dell’accordo di sovvenzione: 956090

DOI

10.3030/956090

Progetto chiuso

Data della firma CE 17 Agosto 2020

Data di avvio 1 Novembre 2020

Data di completamento 30 Aprile 2025

Finanziato da

• EXCELLENT SCIENCE - Marie Skłodowska-Curie Actions

Costo totale

€ 4 095 308,16

Contributo UE

€ 4 095 308,16

Coordinato da

TAMPEREEN KORKEAKOULUSAATIO SR
Finland

Periodic Reporting for period 2 - APROPOS (Approximate Computing for Power and Energy Optimisation)

Periodo di rendicontazione: 2022-11-01 al 2025-04-30

Sintesi del contesto e degli obiettivi generali del progetto

The APROPOS project addresses a growing global concern: the unsustainable energy demands of computing. If left unchecked, by 2040, digital systems could consume more electricity than the planet can provide. This rising energy use, seen clearly in both data centers and mobile broadband networks, has significant implications for climate goals, costs, and long-term digital accessibility. APROPOS offers a novel solution through the use of Approximate Computing (AxC), a technique that intentionally reduces computational precision in areas where perfect accuracy is not necessary. By doing so, it improves energy efficiency by 10 to 50 times without compromising performance. This approach is particularly effective in real-world scenarios like sensor data processing, pattern recognition, and data mining, where "good enough" results are acceptable and expected.

In pursuit of its broader goal to address energy efficiency challenges in future embedded and high-performance computing through disruptive methodologies, APROPOS successfully delivered on its core objectives: equipping researchers with the expertise to earn internationally recognized PhDs, fostering leadership and innovation mindsets through collaboration with SMEs and spin-offs, and enabling the practical application of research in both academic and non-academic sectors. APROPOS also promoted open research practices, strengthened communication and career development skills, supported diversity and inclusion, and helped the fellows make informed choices about academic versus industry careers.

In short, APROPOS successfully trained a new generation of researchers in energy-efficient computing, with a strong emphasis on innovation, open science, and cross-sector and cross-border collaboration. Fellows gained experience in both academic and industrial settings, developed a broad set of transferable skills, and were exposed to diverse career paths and research environments across Europe.

As a result, APROPOS has made a lasting impact on both the scientific community and broader society, offering scalable solutions to reduce digital energy consumption while empowering a skilled, future-ready workforce to carry this mission forward.

Lavoro eseguito dall’inizio del progetto fino alla fine del periodo coperto dalla relazione e principali risultati finora ottenuti

The project saw major advances in both research and training. Fifteen Early Stage Researchers (ESRs) were successfully recruited. By the end of the project, seven had defended their doctoral theses, with the remaining expected to do so within the following year. Collectively, the ESRs produced 69 peer-reviewed publications, contributing new insights and advancing the state-of-the-art in Approximate Computing.

All core objectives of the project were achieved. Strong supervisory teams guided each ESR throughout their research, while training events, delivered in close collaboration with industrial partners, provided both technical and transferable skills e.g. on licensing agreements, technology transfer and grant writing. Intersectoral secondments allowed ESRs to apply their research in diverse environments, and efforts were made to foster an inclusive, multicultural, and gender-balanced research culture. Open science principles were fully embraced, with all results published in Open Access and all ESRs maintaining active ORCID profiles. Additionally, targeted training on fundraising and research proposal writing was provided to support long-term career development.

Public engagement and dissemination have also been central to the project. APROPOS maintained active social media outreach via X (formerly Twitter), Facebook, and LinkedIn, and sustained a regularly updated blog featuring scientific highlights, project news, and outreach activities. ESRs also organized events for schools and the general public, increasing awareness of the importance of sustainable computing and inspiring future generations of scientists and engineers.

Progressi oltre lo stato dell’arte e potenziale impatto previsto (incluso l’impatto socioeconomico e le implicazioni sociali più ampie del progetto fino ad ora)

Beyond its training objectives, the APROPOS project has delivered significant scientific progress, pushing the boundaries of AxC research and application. The ESRs developed and validated a range of algorithms and methods that explore new ways to reduce energy consumption by intentionally trading off computational precision. These contributions address gaps in current research and pave the way for practical, energy-efficient solutions in real-world systems.

Key advancements include comprehensive analyses of existing Machine Learning and Deep Learning techniques relevant to AxC, as well as fault-tolerant approaches and benchmarking tools. Several ESRs explored dynamic and distributed computing strategies, such as computational offloading to network infrastructure and improving satellite-based computing, to further improve energy efficiency. The project also examined the use of AxC in resource-constrained devices like wearables or at the network edge, helping to extend battery life and reduce operational costs. Importantly, some of these innovations have contributed to intellectual property development and technology transfer, strengthening the project's link to real-world exploitation.

By the end of the project, APROPOS delivered a portfolio of openly accessible validated technologies, tools, and insights that can be leveraged by both academia and industry to build more sustainable digital systems. These outcomes have significant socio-economic and environmental implications, helping reduce the carbon footprint of computing, lower energy costs, and support the EU’s green and digital transitions.

APROPOS project logo