WatchEDGE studies an advanced edge-computing architecture supporting AI-based applications distributed on geographically-distant sites. Each site (or “island”) – interconnected by SD-WAN – is equipped with edge-computing infrastructure that can be provided by fleets of flying drones (FANET), and smart radars and fixed or flying cameras.
The project – orchestrated to maximize data processing at the edge – works towards a use case of environmental surveillance for smart agriculture and wildlife protection, relying on AI-based image processing.
WatchEDGE is part of Spoke 8 – Intelligent and Autonomous Systems
Project PI: Guido Maier
The partners also started the activities on "Orchestration Algorithms and Methods" (Work Package 2), currently in preparation and due on M18. The deliverable will describe the project approach to the wildlife monitoring services and related computer-vision and AI methods, to UAV control and SD-WAN algorithms and to the orchestration of the far-edge-egde-Cloud continuum.
The activity of Work Package 3 also started and focused on the planning and design of the POC. In particular, WatchEDGE started an intensive interaction with San Rossore Regional Park in Pisa, where the POC will be hosted. The procedures to buy the experimental equipment by all partners are in progress.
WtachEDGE also disseminated the project goals internally and externally.
They are contained in the two deliverables of the project:
D.1.1. PRELIMINARY SYSTEM ARCHITECTURE AND REQUIREMENT DESCRIPTION
Three main results can be identified:
Both physical architecture and orchestration architecture are original, as they have been developed specifically for WatchEDGE by the cooperation of all the partners.
There is a strong innovation aspect, as the system has been conceived explicitly to provide environmental monitoring services to rural communities. The system is an enabler that allows to bring artificial intelligence to the edge, where it can be applied to many practical functions, such as wildlife control, smart agriculture, fire control, etc.
D.2.1. WATCHEDGE PLATFORM AND ALGORITHMS:EARLY SOFTWARE RELEASE AND REPOERT
[Deliverable due on M18 and still under preparation at the current date (18/06/24)]
The main results achieved in this deliverable are the following::
This deliverable is the first document unrolling the implementation details of the WatchEDGE system. It contains several original scientific contributions on the various aspects of the implementation. The system is an enabler that allows to bring artificial intelligence to the edge, where it can be applied to many practical functions, such as wildlife control, smart agriculture, fire control, etc.
Some of the results mentioned above have been presented in two WatchEDGE dissemination workshops, the first in Catania on September 2023, the second in Pisa on February 2024. The program and presentations of the workshops are available at the link reported below.
WatchEDGE results have also been presented the RESTART plenary meetings in Bari and Bologna, and at other venues.
Further details, including the dissemination workshop programs and presentations, are available at the following link.
Up to this date (June 2024), more than 20 between journal and conference papers, keynotes, invited talks, seminars and tutorials have been produced by WatchEDGE partners and reported in the project. The three most significant papers are the following:
1) Guido Maier, Antonino Albanese, Michele Ciavotta, Nicola Ciulli, Stefano Giordano, Elisa Giusti, Alfredo Salvatore, Giovanni Schembra, "WatchEDGE: Smart Networking for Distributed AI-Based Environmental Control", Computer Networks, 02/2024 [joint paper]
WatchEDGE position paper jointly authored by all the project partners. Provides an overview of the project and of the technical approached that are being and will be followed to accomplish all the scientific goals. The paper also contains an ample discussion about the application scenarios and the real-world use-cases and the expected industrial, societal and economic impact of the project
2) Jean Pierre Asdikian, Mengyao Li, Guido Maier, "Performance evaluation of YOLOv8 and YOLOv9 on custom dataset with color space augmentation for Real-time Wildlife detection at the Edge", ENS Workshop, IEEE NetSoft 2024, 06/2024 [POLIMI]
The paper presents an image data-set enhancement method called color space augmentation, that allows to simulate different lightning scenarios in real-life environments. The method is applied to two deep-learning based image recognition models (YOLOv8 and
YOLOv9). Results obtained by identifying wild-animal images show that the detection accuracy is increased by the augmented color spaces versus the natural ones, as the
proposed technique is able to strengthen the model’s robustness against environmental changes.
As for the innovation aspects, the paper discusses the deployment of these models on far edge devices, such as trap cameras with GPUs, where real-time analysis of wildlife activity
is crucial for both management and conservation efforts.
3) G. M. Cappello, G. Colajanni, P. Daniele, L. Galluccio, C. Grasso, G. Schembra. L. Scrimali, "ODEL: an On-Demand Edge-Learning framework exploiting Flying Ad-hoc NETworks (FANETs)", ACM International Workshop on Recent Trends of Internet of Softwarized Things (IoST-5G&B), co-located with Mobihoc 2023, Washington DC, USA, October 23-26, 2023
This paper proposes ODEL, an On-Demand Edge-Learning framework that uses a Flying Ad-hoc NETwork (FANET) to bring computing and networking facilities on-site for edge learning. ODEL is based on a marketplace employing a non-cooperative game theoretic approach: UAVs are provided by different third-party providers in exchange of some economic gain. A non-linear optimization problem is formulated in order to determine the optimal distribution of flows that maximizes revenue for each UAV provider, and is solved by means of the Variational Inequality (VI) theory.
As for the innovation aspects, the method proposed in this paper allows to overcome a limitation found in many application scenarios commonly envisioned for 6G, i.e. edge learning is not feasible neither locally in the smart objects, due to their computation and energy limitations, nor by servers at the edge of the cabled network, because not connected
with adequate powerful links. The solution proposed will favor edge learning, which brings
machine-learning algorithms at the network edge to achieve massive connectivity, ultra-low latency, energy efficiency, security and privacy. This will provide the society with dynamic
and programmable systems of interconnected smart devices interacting with little to no human intervention.
The adoption of an automated computing/communications/sensing architecture and service automation platform can fulfill different use-cases leveraging on:
Each of the industrial partners involved (4 over 8) has a clear scope and mission within the project, which makes their expertise and contributions well integrated. Contributions of the industrial partners can be summarized as follows:
Cascade calls partners:
The mission of RESTART focused project F1 – WatchEDGE is to offer to the rural community an infrastructure-as-a-service, able to support environmental-surveillance applications based on AI image processing and making use of smart cameras and sensors, both fixed and UAV-mounted.
Community of workers and operators of the rural environment, such as farmers, national-park managers, etc., needs to protect their properties against natural threats, such as wild animals, crop parasites and diseases, raging fires. AI technology is the best allied to effectively control these threats, provided that the community can rely on a powerful and efficient platform able to support AI with enough computing and networking resources.
WatchEDGE studies an advanced edge-computing infrastructure distributed over several geographically-distant sites. It allows raw data to be processed at the edge, while computing functions and trained models are efficiently moved across, thanks to the WatchEDGE SDN-control and orchestration system. Each site is equipped with local edge-computing and networking resources, including fleets of flying drones (FANETs), while inter-site connectivity is provided by SD-WAN.
For further information.
2. Joint Publications
3. Talks/Dissemination events
4. Demo/PoC
5. Project Meetings
6. Patents/Innovations
7. Open source contributions
8. Standardization contributions
Definition of the WatchEDGE physical and control architectures (KPI-1)
Design of the WatchEDGE orchestration system (KPI-2)
Design of the WatchEDGE AI-based application manager (KPI-3)
Development of the WatchEDGE experimental validation testbed (KPI-4)
WatchEDGE-solution performance assessment (KPI-5)
Peer-reviewed publications in top journals/magazines (KPI-6)
Peer-reviewed publications at IEEE/IEEE co-sponsored international conferences (KPI-7)
Fraction of joint publications (e.i. with authors affiliated to multiple RESTART partners (7/18) (KPI-8)
Training events for the scientific community (e.g. conference workshops, keynote, invited or tutorial presentations), and/or for PhD students and/or industry employees (KPI-9)
Tutorial materials (videos, slide sets, etc.) for remote teaching (KPI-10)
International visitors (PhD students/researchers) hosted (KPI-11)
Project Deliverable D1.1: Preliminary system architecture and requirement description [due: M12]
Project Deliverable D1.2: Final system architecture and requirement description [due: M36]
Project Deliverable D2.1: WatchEDGE platform and algorithms: early software release and report [due: M18]
Project Deliverable D2.2: WatchEDGE platform and algorithms: final software release and report [due: M36]
Project Deliverable D3.1: Report on validation and performance assessment [due: M36]
Researchers involved: about 140 person-month (30 univ. permanent staff + 50 univ. RTD/PHD + 60 industry)
Collaboration Proposals:
- WathcEDGE is looking for possible collaboration with end users of the WatchEDGE solution, such as: wildlife parks, agricoltural institutions, farmers, breeders, etc.
- WatchEDGE is contributing to the following RESTART Grand Challenges: Challenge 0 – Create a vision of future evolution of telecommunications ecosystem in Italy and internationally; Challenge 7 – Digitalize the environment for a sustainable world; Challenge 10 – Make the network a platform for programming and running applications; Challenge 12 – Make artificial intelligence distributed and networked; Challenge 19 – Create a community for RESTART open source software
- The WatchEDGE proposal was endorsed by ASSOProvider by a letter of interest
- Within RESTART program, WatchEDGE is seeking co-operation with the structural projects NETWIN, SUPER and COHERENT and with the focused project LEGGERO
For any proposal of collaboration within the project please contact the project PI.
WatchEDGE News