Periodic Reporting for period 1 - OpenSwarm (Orchestration and Programming ENergy-aware and collaborative Swarms With AI-powered Reliable Methods)
Okres sprawozdawczy: 2023-01-01 do 2023-09-30
The OpenSwarm project seeks to revolutionize data-driven systems with low-power wireless technology, focusing on three pillars: efficient networking and management of smart nodes, collaborative energy-aware Artificial Intelligence (AI), and energy-aware swarm programming. Results are consolidated in the open-source software "OpenSwarm," validated on 1,000 node testbeds, and applied to proof-of-concept cases in renewable energy, environmental monitoring, industrial health, and mobility.
Involving nine partners across Europe and running from January 2023 to April 2026, the project aims to achieve five scientific objectives, including improving communication for mobile nodes, enabling energy-aware AI, developing swarm programming frameworks, creating an open-source code base, and validating benefits through real-world applications.
OpenSwarm envisions intelligent nodes contributing to renewable energy, environmental monitoring, industrial health, and mobility. It aims to enhance industrial efficiency, support renewable energy communities, contribute to the European Green Deal, aid biodiversity preservation, and advance scientific progress in distributed AI. The overarching goal is to create energy-efficient, data-centric, and trustworthy systems, transforming various sectors toward a sustainable and collaborative future.
Involving nine partners across Europe and running from January 2023 to April 2026, the project aims to achieve five scientific objectives, including improving communication for mobile nodes, enabling energy-aware AI, developing swarm programming frameworks, creating an open-source code base, and validating benefits through real-world applications.
OpenSwarm envisions intelligent nodes contributing to renewable energy, environmental monitoring, industrial health, and mobility. It aims to enhance industrial efficiency, support renewable energy communities, contribute to the European Green Deal, aid biodiversity preservation, and advance scientific progress in distributed AI. The overarching goal is to create energy-efficient, data-centric, and trustworthy systems, transforming various sectors toward a sustainable and collaborative future.
The summary focuses on WP1-WP4:
WP1 "Requirements and System Architecture": Successfully completed with three main goals achieved in reporting period 1 (RP1): formalizing comprehensive requirements, proposing the initial project architecture, and establishing ethical and safety guidelines. Led by INRIA, the collaboration of partners (ADI, IMEC, ING, SIA, UOS, and WE) refined requirements, designed collaborative AI system architecture, managed proof-of-concept activities, and provided valuable feedback. The successful conclusion establishes a foundation for subsequent project phases, including a comprehensive architecture document, clear technical requirements, and a robust ethical framework.
WP2 "Orchestration of Collaborative Smart Nodes": 25% complete, running until M22, dedicated to developing core technologies for collaborative smart nodes. Contributions from INRIA, ADI, IMEC, ING, KUL, and SIG advance security, firmware management, communication schedules, and innovative approaches to RF technologies and privacy preservation. Progress lays a crucial foundation for OpenSwarm's collaborative smart nodes.
WP3 "Collaborative Energy-Aware AI": 25% complete, running until M22, aims to enable collaborative AI training, inference, and decision-making on low-power swarm devices. INRIA contributes to T3.1 on over-the-air updating, ADI leads T3.1 achieving significant memory reduction, and IMEC leads tasks on online learning, collaborative AI task distribution using Graph Neural Networks (T3.3) and energy-aware collaborative task scheduling (T3.4). Progress aligns with OpenSwarm Project goals for collaborative and energy-efficient AI execution on low-power swarm devices.
WP4 "Energy-Aware Swarm Programming": 25% complete, scheduled until M22, focuses on designing tools for intuitive programming of energy-aware swarm behaviors. ADI, IMEC, KUL, SIG, UOS, and WE contribute to advancing tool design for energy-aware swarm behaviors, energy management, user interfaces, swarm compilers, and virtual machines. Innovations align with project objectives, including hardware, software, and real-world applications.
WP1 "Requirements and System Architecture": Successfully completed with three main goals achieved in reporting period 1 (RP1): formalizing comprehensive requirements, proposing the initial project architecture, and establishing ethical and safety guidelines. Led by INRIA, the collaboration of partners (ADI, IMEC, ING, SIA, UOS, and WE) refined requirements, designed collaborative AI system architecture, managed proof-of-concept activities, and provided valuable feedback. The successful conclusion establishes a foundation for subsequent project phases, including a comprehensive architecture document, clear technical requirements, and a robust ethical framework.
WP2 "Orchestration of Collaborative Smart Nodes": 25% complete, running until M22, dedicated to developing core technologies for collaborative smart nodes. Contributions from INRIA, ADI, IMEC, ING, KUL, and SIG advance security, firmware management, communication schedules, and innovative approaches to RF technologies and privacy preservation. Progress lays a crucial foundation for OpenSwarm's collaborative smart nodes.
WP3 "Collaborative Energy-Aware AI": 25% complete, running until M22, aims to enable collaborative AI training, inference, and decision-making on low-power swarm devices. INRIA contributes to T3.1 on over-the-air updating, ADI leads T3.1 achieving significant memory reduction, and IMEC leads tasks on online learning, collaborative AI task distribution using Graph Neural Networks (T3.3) and energy-aware collaborative task scheduling (T3.4). Progress aligns with OpenSwarm Project goals for collaborative and energy-efficient AI execution on low-power swarm devices.
WP4 "Energy-Aware Swarm Programming": 25% complete, scheduled until M22, focuses on designing tools for intuitive programming of energy-aware swarm behaviors. ADI, IMEC, KUL, SIG, UOS, and WE contribute to advancing tool design for energy-aware swarm behaviors, energy management, user interfaces, swarm compilers, and virtual machines. Innovations align with project objectives, including hardware, software, and real-world applications.
OpenSwarm is making significant progress across its three pillars: collaborative smart node orchestration, collaborative energy-aware AI, and energy-aware swarm programming. Each pillar, represented by work packages around 25% complete, is on track. Notably, one publication is accepted, and four are under review.
In WP2, focused on smart node orchestration, OpenSwarm experiments with advanced network scheduling, standardizes low-power wireless security protocols, adapts Coaty for constrained networks, and translates Zero-Wire communication from visual to RF. A completed Application Performance Monitoring (APM) survey is under review.
WP3, centered on Collaborative Energy-Aware AI, sees OpenSwarm orchestrating complex AI behaviors in intermittently powered swarm networks, utilizing neural network hyper-compression, online learning, and biotechnology-inspired knowledge-sharing.
WP4, Energy-Aware Swarm Programming, unveils innovative methods for programming intelligent device swarms, with two methods set for leading conferences. These methods employ automated evolutionary and heuristic approaches, formal methods, and field calculus for energy-aware and collaborative behavior.
OpenSwarm's scientific leadership is evident in its organization of the Invited Workshop on Crystal-Free Radios, Chip-Scale Wireless Systems, and Swarms in Paris, engaging 26 participants, including 6 from the OpenSwarm ExCom and Scientific Advisory Board. The workshop fosters discussions on the project's scientific program, identifies research topics, and leverages participants as project ambassadors.
In WP2, focused on smart node orchestration, OpenSwarm experiments with advanced network scheduling, standardizes low-power wireless security protocols, adapts Coaty for constrained networks, and translates Zero-Wire communication from visual to RF. A completed Application Performance Monitoring (APM) survey is under review.
WP3, centered on Collaborative Energy-Aware AI, sees OpenSwarm orchestrating complex AI behaviors in intermittently powered swarm networks, utilizing neural network hyper-compression, online learning, and biotechnology-inspired knowledge-sharing.
WP4, Energy-Aware Swarm Programming, unveils innovative methods for programming intelligent device swarms, with two methods set for leading conferences. These methods employ automated evolutionary and heuristic approaches, formal methods, and field calculus for energy-aware and collaborative behavior.
OpenSwarm's scientific leadership is evident in its organization of the Invited Workshop on Crystal-Free Radios, Chip-Scale Wireless Systems, and Swarms in Paris, engaging 26 participants, including 6 from the OpenSwarm ExCom and Scientific Advisory Board. The workshop fosters discussions on the project's scientific program, identifies research topics, and leverages participants as project ambassadors.