Specialist Task Force 584:
Artificial Intelligence for IoT Systems

Who we are:

What we do

The objective of STF 584 is to provide an initially validated architecture that describes how IoT systems can make use of Artificial Intelligence (AI) and Machine Learning (ML) for the management and interpretation of IoT devices data over a large variety of deployment models (e.g., edge or cloud-based) while remaining interoperable, secure and manageable. The challenge is twofold. 

On the one hand, the use of AI and ML techniques are expected to improve drastically the quality of the decisions that can be taken from the analysis of the massive amount of data that is generated by the IoT devices. An example of the new applications and services opportunities offered is predictive maintenance: its goal is to predict component failure and to prevent the occurrence of the failure by performing maintenance. Predictive maintenance brings together various technologies, several of them pertaining to Artificial Intelligence and Machine Learning, such as ad-hoc statistical analysis, predictive modelling, data mining, text analytics, optimization, real-time scoring and machine learning. These techniques support IoT systems administrators by identifying specific patterns in data and make decisions about how to handle what may happen next. The adoption of such AI and ML techniques require the integration of a large number of functions (and associated components) that must be done while preserving the interoperability and openness in the system.

On the other hand, considering that AI and ML are requiring specialised components and intensive computing resources, their usage is associated with additional stringent requirements that have an impact on the IoT system architecture. Consequently, another major objective of this proposal is to identify the requirements for the extension/adaptation of the IoT architectures in order to better integrate AI/ML techniques and ensure that the associated management of data is well handled by the IoT Service Layer.

The rationale for such adaptations will be clarified by an early proof of the validity of the approach based on significant Use Cases and on a prototype implementation of that use case. The purpose of the Proof of Concept activity is to provide a set of results that will serve as an input and an accelerator to further standardization work, in particular through the contributions of SmartM2M to the work of oneM2M.

The applications of AI are concerning a large number of the “verticals” that are addressed by the work of ETSI Technical Bodies and ISGs. Examples of such sectors are the Automotive (with the major case of the ITS Technical Committee) or Smart Cities (with important aspects such as Smart Energy, Smart Buildings or Smart Water Management). Even if those sectors have strong specificities, it is highly possible to outline commonalities on the required evolutions and this should be demonstrated through the analysis of a relevant Use Case that will serve as a support to the analysis and the development of an early validation.

The analysis and early validation of the application is meant to be very focused and as little resource consuming as possible. A major focus of the work will be the contribution to the required evolutions of oneM2M and SAREF and, most importantly, to the required dissemination effort towards the oneM2M and SAREF communities. It is expected that the STF results be provided in a way that will allow oneM2M members to promote their adoption in the oneM2M specifications.

For more details, see our Terms of Reference

Why we do it

Artificial Intelligence (AI) has recently experienced significant breakthroughs, in part due to its ability to build on the promises of both Big Data and Machine Learning (ML) and offer a vast range of promising new services. The applications of Artificial Intelligence technologies are emerging in a large range of Industries and start to provide added value to many ICT systems, including IoT systems. In the vast number of new applications of AI announced daily, still a large number is more expectations than actual implementations based on proven science and technology. In the case of IoT, the situation is not different, and many hurdles still have to be passed before fully catching IA as a business opportunity.

AI is already present in a number of applications in domains such as finance, health, logistics, smart buildings or manufacturing, especially when the deployment of AI algorithms within existing systems is supported by very powerful data management and computing capabilities. In the same movement, it is gradually integrated within of all kinds of networks and support a growing number of deployment models, those cloud-based to start with. There is a potentially large variety of AI applications, but they all have in common the need to be supported by a proper architectural structure, a capacity to easily integrate technology building blocks and components and the need to guaranty continued interoperability.

For oneM2M, Artificial Intelligence comes as a challenge and an opportunity. One the one hand, AI is posing a challenge to oneM2M because of its data-centric approach and its huge requirements in terms of resources available in the cloud domain as well as at the edge of the IoT network. A global approach to the management of data and the deployment of innovative algorithms requires adaptations to the oneM2M architecture that, in any case, already started in order to take into account virtualisation or Big Data. On the other hand, AI is also an opportunity for oneM2M to provide open solutions to applications and services developers together with maintaining and enlarging its core asset of support to interoperability.

How we do it

The STF is integrated within the Work Programme of the SmartM2M Technical Committee (Smart M2M TC).
Apart from the task related to the STF project management (Task 1), the STF works on the development of the deliverables of the STF (i.e. Technical Reports) and on the dissemination of the results towards the oneM2M community within the following tasks:

• T2        Identification of Architecture Evolutions
• T3        Proof of Concept
• T4        Dissemination towards oneM2M and the IoT community 

Deliverables

• ETSI TR 103 674: SmartM2M; Artificial Intelligence and the oneM2M architecture
• ETSI TR 103 675: SmartM2M; AI for IoT: A Proof of Concept

Time plan

The work of the STF has started on January 6th, 2020 and will be finished by the end of December 2020.  The main milestones of the project are listed in the table below:

How to contact us

Please contact us via email: emmanuel.darmois@commledge.com