Skip to main content
European Commission logo
italiano italiano
CORDIS - Risultati della ricerca dell’UE
CORDIS
CORDIS Web 30th anniversary CORDIS Web 30th anniversary

Very Efficient Deep Learning in IOT

Periodic Reporting for period 2 - VEDLIoT (Very Efficient Deep Learning in IOT)

Periodo di rendicontazione: 2022-05-01 al 2024-01-31

The VEDLIoT project targets the development of energy-efficient Deep Learning for distributed AIoT applications. A holistic approach is used to optimise algorithms while also dealing with safety and security challenges. The approach is based on a modular and scalable cognitive IoT hardware platform. Using modular microserver technology enables the user to configure the hardware to satisfy a wide range of applications. VEDLIoT offers a complete design flow for Next-Generation IoT devices required for collaboratively solving complex Deep Learning applications across distributed systems. The methods are tested on various use-cases ranging from Smart Home to Automotive and Industrial IoT appliances.
The VEDLIoT initiative has successfully concluded, achieving its core objectives and making substantial contributions to the evolution of Cognitive IoT. Key outputs from the project include the development of a flexible Cognitive IoT Hardware Platform, a multi-tier Reconfigurable System Architecture, and innovations in Co-Design for Adaptive Hardware Accelerators that have significantly improved resource utilization. Efforts also led to notable enhancements in the energy efficiency and performance of IoT systems.
Further accomplishments encompass the development of a powerful Toolchain for Distributed AI, the introduction of comprehensive security, privacy, and trust frameworks, and advancements in durability and functional safety, especially within AIoT and automotive sectors. The project showcased broad applicability, deploying VEDLIoT technologies in diverse fields such as industrial IoT, automotive technology, and smart home applications.
Teamwork and collaboration were pivotal to VEDLIoT's achievements, with effective integration and synergy between various teams and project components fostering cohesive solutions and steady progress. VEDLIoT was proactive in sharing its breakthroughs and insights by participating in key industry gatherings and forming partnerships, thereby promoting exchange of knowledge and broader impact.
In essence, VEDLIoT's comprehensive strategy in technological innovation, collaboration, communication, and application has significantly propelled the field of Cognitive IoT forward, establishing new benchmarks. VEDLIoT has resulted in 21 novel exploitation outcomes, spanning software and hardware, and incorporating both open source and proprietary components. Moreover, VEDLIoT's contributions have been widely shared through 15 journal articles, 33 conference presentations, and visibility at 11 exhibitions featuring a dedicated VEDLIoT booth.
The VEDLIoT project has yielded several critical insights essential for the creation and deployment of AIoT systems. A key takeaway is the necessity of diversity and adaptability in hardware designs. Tailoring hardware to suit the unique demands of various applications can lead to enhanced performance and energy savings. Employing a modular and scalable design, based on standard industry formats, facilitates support for a wide array of applications in different fields.
The importance of comprehensive toolchain support is another vital insight. The facility to efficiently translate Deep Learning algorithms across all principal processing structures within adaptable and varied hardware proves to be of great value. This feature not only capitalizes on the capabilities of sophisticated hardware but also aids in minimizing energy use and prolonging the operational life of distributed AIoT systems, thereby promoting sustainability.
Security and durability have also been identified as crucial components. The implementation of distributed attestation mechanisms and secure execution environments is fundamental for a wide range of AIoT applications. Together with initiatives aimed at ensuring system robustness and safety, these security measures facilitate the use of AI in critical systems, elevating their reliability and credibility.
Moreover, the systematic engineering of requirements for AIoT stands out as imperative. Adopting a structured framework for the meticulous and inclusive planning of distributed AIoT systems guarantees the fulfillment of all specifications while adhering to legal standards, like the AI Act. This methodical strategy plays a pivotal role in developing AIoT systems that are both efficient and in line with current and forthcoming regulations.
Smart home use case
VEDLIoT Toolchain
Hardware - u.RECS
Automotive use case
Hardware - t.RECS
Hardware - RECS Box Deneb
Hardware - RECS Box Durin
Projet Logo