MIRAI | Machine intelligence techniques for smart and sustainable planning and operation of IoT and Edge computing applications

For the last two decades the standard approach for IoT applications has been to leverage cloud infrastructure to overcome constraints on the level of end and edge nodes. Especially with the advent of applications with hard, real-time requirements and increasing AI usage and its demand for more storage and computational power, this approach alone is not viable any longer. MIRAI aims to develop the MIRAI Framework Building Blocks (MFBB), based on distributed AI techniques, to enable smart and sustainable planning and operation of IoT and edge computing applications. This will supplement the traditional cloud-centred approach with horizontal scaling amongst distributed edge devices, such that the full richness of the data can be leveraged without bandwidth constraints.

Context

Only 1% of the data generated by end nodes and available at the edges of modern networks is utilised; the rest is neglected due to limitations, such as low bandwidth and high latency in the connection to the cloud. Furthermore, for most IoT devices the security and privacy standards are still poor. The current approach for IoT is to leverage cloud infrastructure to address constraints at the end/edge nodes, but this is no longer viable due to hard, real-time requirements of (mission-) critical applications, increasing AI usage and high demands on storage and computational power.

As a decentralised intelligence framework, MIRAI will enable the optimal distribution of AI computing tasks and workloads across existing computing nodes, serving as a truly scalable edge computing software toolkit for IoT and edge computing applications. Through the MIRAI Framework Building Blocks (MFBB), appropriately sized AI modules will be deployed at nearby available edge nodes. This will provide a low-latency distributed ecosystem for AI-enabled computing in IoT. With application services and tasks deployed on local resources, network problems will become less critical. This distributed approach will make the MIRAI solution more robust - by enabling new failover mechanisms-  and secure as the computations are executed directly on the source without the need to move possibly private and security-critical data around.

In Belgium, the MIRAI solution will be applied in the domains of distributed renewable energy systems, traffic management and water management.

Objective and results

• Within the Belgian consortium, the focus is on three different use cases:
  • The leading-edge technology company and expert renewable energy consultant 3E aims at improving the real-time control of renewable energy assets by automatically reacting to the information available at edge level, without the need of offloading data to the cloud. 
  • As a leader in smart mobility solutions Macq aims at using the data obtained by their traffic cameras to make the streets safer for vulnerable road users, such as pedestrians and bikers, e.g. by combining information from different cameras at the same crossroads, without the need of saving and sharing this highly private data.
  • The Certified B Corp Shayp intends to reduce the waste of fresh water by putting parts of the leakage detection intelligence to the edge level, with the result of quicker detection of possible leakages, but also with the possibility to reduce the messaging bandwidth between the edge and the central device.
• MIRAI will provide publicly available knowledge to the research and industrial community, e.g. about MIRAI ML/AI algorithms and data-mining models specification and design as and data sharing policies & processes between edge devices.

Funding

• Funding agency/modality: 
• ITEA framework with support from Innoviris.