Mobile communications have pervaded our everyday lives, deeply affecting the way in which we work, travel and spend our free time. We use daily mobile applications and services and fifth generation (5G) and beyond 5G networks are expected to lead to a further digital revolution, enabling ubiquitous and enhanced broadband services, smart/autonomous vehicles, intelligent transport, and complex human machine interactions (e.g., industrial IoT and extended reality).
That is, our connected society will generate a vast amount of data, which has to be transmitted, processed and analysed, often in an online fashion. Especially, this data surge demands a massive use of computing services based on artificial intelligence (AI) and machine learning (ML) algorithms. With the advent of 5G networks, these algorithms will be increasingly pushed towards the edge of the mobile wireless network and, at a global scale, they will have a large, unsustainable, carbon footprint.
The Greenedge mission
The Greenedge project’s mission is to tame the growing carbon footprint of ML/AI based edge computing services adopting a twofold approach.
- As a first step, renewable energy resources, such as wind or solar power, will be utilized to sustain the communication and computing tasks that run at the mobile network edge.
- Second, fundamentally new and green-by-design computing and communication paradigms will be developed, to wisely exploit the available communication, computing and energy resources across the mobile edge network. This entails the design of novel ML/AI in-network processing techniques that have a small memory footprint, and that are at the same time trainable in an energy efficient manner. Particular attention will be paid to distributed computing approaches (federated learning).
This will be accomplished by creating a training network of 15 early stage researchers, who will carry out research on Greenedge sustainable computing paradigms. They will confront a highly cross-disciplinary endeavour, mastering the fundamentals of energy generation and storage, mobile network management, machine learning/neural networks, mathematical optimisation and distributed computing.
These young researchers will be trained via specialised training schools and supervised by leading experts in their fields. By the end of the project, they will be awarded a Ph.D., and will be ready to take a prominent role in the industry or academia, leading the design and implementation of green computing systems for the years to come.
Moving towards green mobile networks
At Greenedge, we believe that a sustainable development is crucial to promote prosperity while protecting the planet, as stated in the United Nations (UN) Sustainable Development (SD) agenda. In fact, ICT is a fast growing sector that benefits societal prosperity by opening new markets and opportunities for both individuals and companies. However, often the sustainability issues that arise from the deployment of this technology (e.g., the carbon footprint) are not taken into account at the design stage. Instead, according to the UNESCO Engineering Initiative, all engineering fields should incorporate sustainability into their practice in order to improve the quality of life for all citizens, and not leaving it as the sole responsibility of environmental engineering. Greenedge follows the guidelines for a sustainable design in ICT and specifically targets UN SD Goal 7 “Affordable and Clean Energy” and Goal 13 “Climate Action” by taming the environmental impact of ICT through the use of green edge computing platforms. Moreover, Greenedge will contribute to Goal 9 “Industries, Innovation and Infrastructure” and Goal 11 “Sustainable Cities and Communities” though our studies on how MEC can serve vertical industries. Finally, the implementation of the project may enable Goal 17 “Partnership for the Goals”, by establishing closer relations among the partners, including the private sector, public organisations, governmental bodies and civil society that will support our project and will be invited to our events.
There is still a significant skills gap in the area of sustainable design in ICT. GREENEDGE is a valuable opportunity for training new young talents, who will face technical, analytical and strategic problems for the sustainable design of MEC platforms to respond to future societal challenges and needs.
Greenedge scientific objectives
- Obj1 (ESR1-4, ESR6-7, ESR9) To explore optimisation and machine learning algorithms for distributed and online resource scheduling in extreme environments with intermittent connectivity (unreliable links) and scarce energy availability.
- Obj2 (ESR12, ESR13, ESR14, ESR15) To explore emerging software frameworks and hardware architectures for distributed and energy efficient edge computing
- Obj3 (ESR5, ESR8, ESR10, ESR11) To profile energy consumption and memory footprint of ML algorithms on selected hardware architectures.
- Obj4 (ESR1, ESR4) To develop lightweight and fast algorithms for the elastic, energy efficient and online orchestration of computing resources in energy harvesting MEC networks
- Obj5 (ESR4, ESR6) To develop advanced decentralised training of ML models (e.g., federated learning) for MEC in the presence of energy, memory and link reliability constraints.
- Obj6 (ESR3) To activate and place data-plane (user/field data) functions in the wireless edge via neuro-inspired unsupervised algorithms, towards enabling a highly energy efficient edge management.
- Obj7 (ESR2, ESR7, ESR12, ESR14) To design and implement cross-layer and end-to-end learning architectures and algorithms for the joint orchestration of communication and computing in energy harvesting MEC networks.
- Obj8 (ESR8) To devise AI-based algorithms for energy consumption profiling and the semi-supervised detection and classification of energy consumption anomalies in the wireless edge.
- Obj9 (ESR11) To devise energy efficient and distributed techniques for the detection of security threats in wireless edge networks.
- Obj10 (ESR5) To develop multi-data source and energy efficient edge sensing, data fusion and computing means, to provide human density/mobility estimation and anomaly detection services in urban environments (smart cities).
- Obj11 (ESR9, ESR13) To design and implement energy efficient communication and edge computing techniques for critical infrastructures and extreme environments with intermittent connectivity and scarce energy availability, under high reliability and low latency requirements.
- Obj12 (ESR10, ESR15) To design and implement near zero power, edge communication (light-based IoT (LIoT)), computing, and sensing technology for smart indoor spaces.