6G Bridge - Liquid AI for 6G software JYU (LiquidAI)
Päärahoittaja
Rahoittajan antama koodi/diaarinumero: 8542/31/2022
Päärahoittajan myöntämä tuki (€)
- 420 000,00
Rahoitusohjelma
Hankkeen aikataulu
Hankkeen aloituspäivämäärä: 01.01.2023
Hankkeen päättymispäivämäärä: 31.12.2025
Tiivistelmä
A typical Internet of Things (IoT) as well as the proposed 6G network system consists of a large number of different subsystems and devices, including sensors and actuators, gateways that connect them to the Internet, cloud services, end-user applications and analytics. Today, these subsystems are typically implemented with a broad variety of programming technologies and tools, making it difficult to migrate functionality from one subsystem to another.
At the same time, applications - especially those that build on machine learning - include and generate huge amounts of data. Although all devices are assumed to be connected, all data does not necessarily move because of privacy or performance reasons. Thus, it is necessary to bring computing close to data.
In this project, we propose ‘liquid’ IoT system architectures, where software and applications can flow from one device to another. Such liquid systems are enabled using a consistent set of technologies in all the subsystems, thus allowing different parts of the system to run the same code. In this project, we apply the technologies to AI subsystems, which have so far been under the radar in our research.
The promise of the project is savings in the implementation work for the different pieces of software by allowing liberal deployment architectures of intelligent functions. Depending on the complexity of the application, the cost of reimplementing already existing functions for certain contexts can be considerable, and it will be accumulated by the fact that software also requires maintenance and upgrades. This gives a considerable upper hand for companies that can deliver such flexible solutions globally.
At the same time, applications - especially those that build on machine learning - include and generate huge amounts of data. Although all devices are assumed to be connected, all data does not necessarily move because of privacy or performance reasons. Thus, it is necessary to bring computing close to data.
In this project, we propose ‘liquid’ IoT system architectures, where software and applications can flow from one device to another. Such liquid systems are enabled using a consistent set of technologies in all the subsystems, thus allowing different parts of the system to run the same code. In this project, we apply the technologies to AI subsystems, which have so far been under the radar in our research.
The promise of the project is savings in the implementation work for the different pieces of software by allowing liberal deployment architectures of intelligent functions. Depending on the complexity of the application, the cost of reimplementing already existing functions for certain contexts can be considerable, and it will be accumulated by the fact that software also requires maintenance and upgrades. This gives a considerable upper hand for companies that can deliver such flexible solutions globally.
Vastuullinen johtaja
Päävastuullinen yksikkö
Liittyvät julkaisut ja muut tuotokset
- LiquidAI : Towards an Isomorphic AI/ML System Architecture for the Cloud-Edge Continuum (2023) Systä, Kari; et al.; A4; OA; 978-3-031-34444-2
- Towards Liquid AI in IoT with WebAssembly : A Prototype Implementation (2023) Kotilainen, Pyry; et al.; A4; OA; 978-3-031-39764-6
- WebAssembly in IoT : Beyond Toy Examples (2023) Kotilainen, Pyry; et al.; A4; OA; 978-3-031-34444-2
