11 KiB
EMELIOT will contribute to research and industrial practice by conceiving and assessing innovative solutions for developing, testing, deploying, and operating dependable ML-based IoT systems. In the following, the EMELIOT project's impact is discussed under different dimensions, including its potential economic and industrial impacts, and the potential key roles it can play to support the protection of cultural heritage and of the natural environment.
The Internet of Things is rapidly gaining ground in different application domains. According to the US National Intelligence Council, \`\`by 2025 Internet nodes may reside in everyday things - food packages, furniture, paper documents, and more'' (NIC, 2008). According to a recent report by IoT analytics (IoT-Analytics, 2020), in 2020 IoT connections (i.e., those established by connected cars, smart home devices, industrial equipment, etc.) surpassed non-IoT connections for the first time. According to the same IoT analytics report, by 2025 the expected number of IoT connections will be more than 30 billion, i.e., four IoT devices per person on average. With the advent of new technology standards like 5G, IoT devices are expected to grow much faster.
ECONOMIC IMPACTS
The availability of a massive amount of data from multiple sources is pushing business across the globe to collect and analyze such data. Under such circumstances, according to a recent report (GAI, 2020), Artificial Intelligence and Machine Learning are gaining a key role in IoT enabling the automated identification of patterns, detection of anomalies, and in general they will permit different stakeholders to gain insights from the huge amount of available data. In this respect, the global AI and IoT market size is expected to grow from $6.81 billion in 2020 to $18.2 billion in 2025 (GAI, 2020). Moreover, according to the recent Global Industry Analysts, Inc report (Research Markets, 2020) the global market on Machine Learning is expected to reach $27.7 Billion by 2027 growing at a rate of 38.4% over the analysis period 2020-2027. However, according to GAI (2020) the lack of a skilled workforce, together with concerns related to data security, are some of the major factors that can restrain the expected market growth.
EMELIOT will define techniques and tools to support the whole life-cycle of ML-based IoT systems. The envisioned platform will permit teams consisting of different figures with heterogeneous expertise (e.g., data scientists, ML experts, and software engineers) to collaborate on the development and management of complex ML-based systems. The whole software life-cycle will be guided and supported by leveraging MLOps practices. Development environments and recommendation systems will be provided to support the early phases of system development; advanced testing and monitoring techniques will be devised to check the dependability of the deployed systems. Concerning security aspects, adversarial machine learning techniques will be conceived to avoid malfunction in developed machine learning models.
Thus, the EMELIOT adopters will benefit novel technologies to enter a promising market or even make their presence more robust. The IoT market is growing very fast, and it necessitates advanced approaches supporting the development of dependable ML-based IoT systems. In such a context, the availability of the right tools and technologies is crucial for ensuring economic growth in a very competitive context.
TECHNOLOGICAL INNOVATIONS AND IMPACTS ON INDUSTRIAL APPLICATIONS
The adoption of the Internet-of-Things is significantly changing the way industries work. For instance, manufacturing processes involve heterogeneous interconnected smart devices, which collaboratively work on core production operations to optimise them and disclose further revenue possibilities. Digital transformation can be fully achieved only if advanced technologies are exploited to take full advantage of large amounts of data, which are collected by different and heterogeneous sources including employed industrial robots, sensors used to monitor various smart buildings, and devices that monitor at real-time the overall system effectiveness.
Traditionally, the management and analysis of collected data are performed by cloud services. However, according to a recent GlobeNewsire report (ECM, 2020), the scenario is changing with the increasing spread of novel technologies, like sensors and other data-producing devices, which permit to perform also analysis tasks in the network edge. According to the same report, the edge computing market is expected to grow 29.4% by 2025. To this end, dedicated techniques and tools are needed to improve the management and analysis of larger datasets in a distributed manner. EMELIOT is expected to provide relevant contributions in this respect by devising novel architectures underpinning federated and distributed learning techniques. They will permit to optimise ML training activities by mixing cloud and on-premise services by adequately distributing training tasks on different platforms. The exploitation of edge computing is considered as a critical factor to foster the market growth predicted by the marketdataforecast report (GAI, 2020).
Overall, EMELIOT will foster dependable software solutions, which can enhance industries under different perspectives. In particular, ML-based IoT systems developed, verified, and maintained with EMELIOT technologies will improve industrial applications by optimizing maintenance, reducing costs, avoiding congestion and hazards, and improving sustainability.
CULTURAL HERITAGE AND ENVIRONMENT
Over the last years, several IT solutions have been proposed for preserving the environment and Cultural Heritage (CH) sites. The availability of devices characterized by low-power consumption, extensive connectivity, and small size, fostered the possibility of continuously monitoring monuments, art exhibitions, environmental parameters, etc., for instance, to fight vandalism, to sense structural changes in historic buildings, and to protect vulnerable natural environments. Machine learning techniques can play a crucial role in such contexts, e.g., to predict structural issues of monitored facilities, which might require dedicated maintenance operations. Moreover, they can anticipate dangerous situations, e.g., due to high air pollutant concentrations in some of the monitored areas. Unfortunately, analyzing the vast amount of data produced by the monitored sites can be difficult and require advanced expertise. EMELIOT will be a precious support to develop this kind of software systems. In particular, the EMELIOT technologies will simplify the collaboration of different and heterogeneous professionals, including software developers, data and climate scientists, and CH experts that might need working together to protect CH sites of interest or specific natural environments.
SCIENTIFIC AND TECHNOLOGICAL IMPACTS
The scientific and technological impact of the project will be based on top-level publications and demonstrations, and it will be articulated in many directions:
- A novel low-code environment to model ML-Intensive IoT systems in a systematic way;
- Newfangled recommendation systems specifically conceived to support different stakeholders, all involved in the development of ML-Intensive IoT systems. Interestingly, the envisioned tools will be able to manage in a homogeneous manner different kinds of artifacts ranging from e.g., data collected from IoT sensors, to snippets of source code that are relevant for the current development task;
- Innovative approaches and technologies to support the validation and verification of specified ML-Intensive IoT systems;
- Innovative techniques and tools supporting the continuous integration, delivery and monitoring of ML-Intensive IoT systems.
- A novel MLOps infrastructure supporting all the aspects related to the development and operations of ML-Intensive IoT systems. The infrastructure will foster a systematic collaboration and interaction of different professionals involved in the development of this kind of systems including domain experts, data scientists, ML experts, and software developers.
DISSEMINATION ACTIONS
The EMELIOT partners have identified three communities to foster the use of the envisioned techniques and tools developed in the project:
- Software developers of IoT systems will create demand for the project technologies as they become aware of the substantial improvements EMELIOT provides for developing ML-intensive IoT systems. Specific dissemination actions: The project will carry out a wide range of dissemination actions to create awareness of the project results amongst Software Developers of IoT systems. Industrial and tutorial tracks of scientific and technical events will be targeted to organize demonstration and hands-on sessions where EMELIOT technologies will be shown in practice.
- System Integrators are a separate community targeted by EMELIOT as they often deliver IoT solutions for enterprise and government. They often have the capabilities to adopt open source technologies and utilise them within the commercial applications they deliver. Specific dissemination actions: The project partners will target System Integrators in dissemination actions and encourage their evaluation and eventual adoption of the EMELIOT technologies for custom application development. All the EMELIOT partners have in place several collaborations with National and International players developing complex software systems for their customers. Technical and demonstration sessions will be organized with them to show the technical results achieved in the project and the potential benefits related to the adoption of the EMELIOT technologies.
- Academia and Researchers communities are important because they can contribute to the evolution of the EMELIOT technologies via a sustainable open source process. Dissemination actions will target technical journals, conferences and other academic and research events and communications channels. Specific dissemination actions:There are several different communities that will be highly impacted by the project results, including general communities of practice in software product development, machine learning, software product life-cycle management as well as IoT system modelling and analysis. The relevant research communities maintain a large number of international journals and conferences. Only very few examples of relevant scientific journals that will be targeted by the project for papers can be listed such as: IEEE Transactions on Software Engineering (IEEE), Software and Systems Modeling (Springer), Empirical Software Engineering (Springer), International Conference on Software Engineering (ICSE), International Conference on Model-Driven Engineering Languages & Systems (MODELS), International Conference on Software Maintenance and Evolution (ICSME), ACM Conference Series on Recommender Systems (RecSys), International Conference on Machine Learning (ICML), International Conference on the Internet of Things (IoT)
The partners have the capabilities and existing contacts to address each of the three target markets for the project results.