top of page

H2TRAIN

by Juan Antonio Montiel Nelson and Marco Ottella




In a Eurostat article on population structure and ageing, with data extracted in February 2024, an overview of the impact of demographic ageing within the European Union (EU) claims that “probably the most important change will be the marked transition towards a much older population structure, a development that is already apparent in several EU Member States.”


Ageing is, of course, an unavoidable process and the progress of medicine as well as the improvement in conditions of life (including lower mortality from illnesses) has given the possibility to live longer. However, longer does not mean ‘quality’. Diseases like dementia and diabetes type 2, as well as cancer to some extent, have a strong correlation with ageing and with the progressive deterioration of the DNA-replication mechanism. However, a healthy lifestyle in terms of regular movement and a balanced diet has been proven to extend quality life. However, stress and other factors can hinder the purpose where we, as humans, also need motivation and support to move around properly to avoid injuries (you don’t run a marathon starting from the couch) and so on. The H2TRAIN idea is to advance the available technologies (sensors, AI, edgecloud continuum…) and integrate them smartly up to a higher TRL to achieve an ecosystem for remote coaching.


Similarly, in the Healthcare sector, in the area of rehabilitation of patients with chronic diseases or post-surgery treatment, we found a high technology demand for remote monitoring (home hospitalisation). Undoubtedly, these two technologically greedy topics demand impactful applications in terms of economic and societal value whereas wearables technology has grown dramatically in the area of sport and physical recreation. From our perspective, a technology gap exists between both kinds of applications, one that H2TRAIN will reduce.


Health, wellbeing and the digital society 

H2TRAIN is an innovation action, more precisely, a Joint Undertaking Innovation Action under the Horizon Europe Framework Programme, in the Non-Initiative (former KDT) of the Chips JU. The research and innovation in H2TRAIN address the enabling digital technologies in holistic health-lifestyle supported by artificial intelligence (AI) networks. H2TRAIN is focused on the major challenges and priorities of the Strategic Research and Innovation Agenda for the Electronic Components and Systems of the European industry: health and wellbeing and digital society. In H2TRAIN a number of advanced wearable devices and biosensors are integrated into textiles for monitoring physiological signals, including biomarkers. The physiological signals are pre-processed by AI on the edge, and by digital twins in the cloud while health or sport experts monitor the evolution of individuals.


Many technologies are already available in the form of chips or in the form of algorithms. H2TRAIN intends to bring them to a higher level of maturity, leveraging on the strength of the European industry to tackle the weakness of the application part of the supply chain. In the meanwhile, we aim to create a platform, an ecosystem also in the sense of cluster or projects (for instance, through a specific focus topic) to put this on the same table with other actors (industry, healthcare institutions, sports teams) and projects (Chips JU but also HORIZON as well as other national and transnational initiatives – for instance, THCS) and work together towards the same objective. Make Europe a leader, or at least recoup part of the terrain that has been lost in these applications which are of strategic importance for our future.



The power of collaboration 

The Horizon magazine (https://projects. research-and-innovation.ec.europa.eu/en/ horizon-magazine), is a comprehensive site containing plenty of information on several topics and Horizon projects, but under the topic of health there is no action regarding the application of AI to health and wellbeing. Nor are there any innovation actions for elderly people or the rehabilitation of patients with chronic diseases. These are two areas that demand major impact in terms of economic and societal value.


Research cannot be done alone with only the technical experts. Other competencies (the so-called non-STEM, for instance, SSH) must be brought to the table to support the vision. Stakeholders as well as end users must be involved from the very beginning and during the development process as living KPIs that monitor the execution and the deviations from the original purpose and ensure that systems are made affordable, useful, easy to use and our lives better, healthier, more fun.


The brokerage as a springboard 

Marco Ottella looks back at where it all began. “At the 2022 brokerage, the first face-to-face event after COVID, thanks to preliminary discussion I had with experts in sports sciences and in artificial intelligence, I conceived a project idea to bring to a higher level the wearable sensors that many of us hobby sporters use every day in order to address the possibility to better monitor sports activities both from professional trainers and from AI-based virtual trainers, with the evident benefits for society as discussed above (quality of life, reduced costs for healthcare). At the brokerage, I had the good fortune to meet some representatives from the semiconductor industry with whom we discussed the weakness (costs) and the opportunities (multimarket) of such an approach. Interest at the presentation and at the poster sessions was big. More than 100 business cards were collected and same number of emails received in the following days. As I said to all of them, I was ready to write the project from A to Z, but with Xtremion being a startup, I could not be eligible as a coordinator. Many of the potential partners who had said they were interested also said that they could not be a coordinator. Luckily, I met Juan at the Brokerage, and a few weeks later I came back to him to ask if he would be happy to coordinate. Thanks to his superior capacities of coordination and my ideas as basic ingredients, he was able to write an outstanding project outline which was evaluated even better than expected.” The rest is history. “The FPP was excellent,” Marco continues, “and we scored 14.5 points out of 15, something that had never happened to me in the last 15 years. One of the best coordinators I have ever seen: not just asking partners for inputs (and many times neglecting them) but being positive everyday with well written and sound text, charts, figures, tables…. A very fruitful cooperation which I hope will last a long time.”


Juan goes on: “While there is no unique method for building a consortium, the ECS Brokerage Event organised by the industry associations AENEAS, EPoSS and INSIDE, is an efficient tool for creating a group of interest around an idea. By promoting an idea as a poster, you will discover how much interest it creates. The project idea presentation is a perfect opportunity for partner finding and, finally, face-to-face consortia building meetings help build the structure of your future proposal. I have attended ECS Brokerage each year, after the pandemic, and promoted several ideas in the area of graphene sensors and thermoelectricity. In 2023, I promoted a biomarker-sensing wearable for stress monitoring. I had previously written a proposal of 45 pages for a Horizon 2020 RIA call and a poster for presenting the idea behind it. During the brokerage the high number of business cards on my poster area and the received emails from potential members suggested to me to prepare a preliminary meeting. After the event, I met Marco and through two video conferences, we expanded the original ideas, from foundational to cross-sectional technologies, incorporating AI/ML from the edge to the cloud. In two weeks, I wrote the proposal and provided the draft to Marco for his revision, and we began to scale up for building the consortium. Marco is the best technical manager, his broader view and industrial experience are the key factors for a winning proposal. One year later, here we are, continuing to work at the kick-off meeting of H2TRAIN. There is more work behind project consortium building, of course, but the ECS Brokerage Event was a good start.”


Building an innovation action plan 

H2TRAIN is focused on the major challenges and priorities of the ECS-SRIA 2023 and 2024 in the application areas of health and wellbeing and digital society. The project aims to unleash the innovation potential of digital technologies onto the smart healthcare market with a focus on the elderly, rehab of chronic disease patients and recreational sports activities. H2TRAIN IA is spanning the entire electronics components and systems, and systems of systems, value chain, from foundational and cross-sectional technologies to the application fields of remote assisted-living, for elderly people, patient rehab monitoring and recreational sports training.


In technology, most projects build from bottom to top. While this approach has its advantages and disadvantages, the major handicap concerns defining the application in terms of the technology, and this is a big issue because it reduces the freedom of degrees when you approach the top. However, when moving from top to bottom, most concepts at application level have no limitations. The problems begin when the foundational technology is not available, but this is not the case. From top to bottom, a number of technology demonstrators are defined, like a tattoo sweat-sensing device, a glycaemic instrument, a pH device based on printed electrodes, ECG, EMG and SpO2 on-textile device; the cortisol, lactate and C-reactive protein based on graphene; the in-water activity tracker; the energy harvester based on thermoelectricity, RF and magnetomechanical; the biometric cryptography device; the textile activity tracker, and the edge-cloud AI continuum processor. But as cross-sectional technologies, there is a huge number of software layers for implementing the functionalities at the edge such as plug-and-play, smart boot for software loading and update, IP licensing, fog AI/ML computing, smart communication interface, embedded intelligence (both supervised and unsupervised operation modes) and advanced embedded intelligence.


The semiconductor industry, with its enormous complexity, suffers from a major structural issue, where investments for fabrication (usually amounting to billions of euros) can be justified if, and only if, the volumes are in the order of several millions per year. This makes the healthcare sector a very difficult one for semiconductor manufacturers. Although miniaturisation can unleash immeasurable benefits (think for instance of miniaturised dispensers for drugs, or blood/heart screening, as well as nanoscale x-ray machines and so on), they are simply not available because they are not affordable. The key is reuse: using the same hardware in several sectors (including consumer) and let the advanced features be implemented as software or ML algorithms.


Innovating the applications 

In H2TRAIN, an interoperative hardware/ software layer between the application level and the technology demonstrators provides the transition between edge and cloud worlds, as an edge-cloud AI continuum. Technology demonstrators have already been introduced. The application level is a common infrastructure for each use case: Remote Assisted Living (RAL), Intelligent Adaptive Sport Coaching (IASC) and Remote Post-Surgery & Rehab Monitoring (RPS&RM). The application is a set of hardware/ software development based on computer desktop technology for: 1) information and communication technology, abbreviated as ICT; 2) ambience monitoring technology, abbreviated as Ambience; 3) digital twin technology; as Digital Twin 4) expert centre technology, as Expert Centre; 5) individual communication gadget, as Individual; 6) individual related communication gadget, as Individual Related.


All these applications are available through cloud-computing networks as against edgecomputing as is the case with technology demonstrators. It should be noted that a particular technology demonstrator, named TD10, acts as an interoperative hardware/ software layer between the application level and the rest of technology demonstrators. For example, TD10 incorporates smart switching capabilities for quality of service when the rest of technology demonstrators deliver data to the main servers of the cloud-computing network. In addition, TD10 provides extra capabilities for AI/ML algorithms more beyond the embedded intelligence layer of the technology demonstrators. Data compression is also a characteristic of the TD10 layer, where data compression standards are applied for reducing the amount of network-traffic load. Plug-and-Play (PnP) operation is included in TD10 for enabling the system to adapt to hardware changes with minimal intervention by the user. A user can add and remove devices without having to do manual configuration, and without knowledge of computer hardware. An Intellectual Property (IP) security layer is provided by TD10, by checking the hardware/ software licence both locally and in remote licence servers. Similar to system integration of human-centric technology and embedded intelligence, here an embedded intelligence system is being developed based on an ultra-low power microcontroller for pushing foundational technologies toward Artificial Intelligence & IoT or Artificial Intelligence of Thing (AIoT) applications in health and sport. Therefore, this goal is also geared towards the development of AIoT devices, bringing new opportunities for sensor data management, as applications and services are moving developments towards the edge instead of in the traditional centralized cloud data centre. TD10 is aimed at the implementation of sensor fusion functionalities, using microcontrollers based on processor architectures such as ARM Cortex Mi (i = 0, …, 4) or recently RISC-V.


From cross-sectional technologies to fog layers 

AI/ML are technologies that flood everything nowadays, but at application level, and not at technology level. H2TRAIN goes from application to foundational technologies. We use digital twins for performance estimation when individual monitoring, and the acquired information is very useful for an expert, which is the medical centre in charge of the supervision for a patient or an elderly person or even the sports coach. Wearable devices and IoT are the media for supplying the information, but the cross-sectional technologies play a key role; where some AI/ML processes run on the edge in comparison with the vast majority of AI/ML algorithms that are executed in the cloud. This is exactly what we incorporate as innovation to overcome the limitations of wearable technologies, which are limited in terms of energy and power consumption, an issue that is very well known for sports practitioners. We found that a proper distribution of the AI between edge and cloud will relax the energy consumption of data communication.


In H2TRAIN, an interoperative hardware/ software layer between the application level and the technology demonstrators provides the transition between edge and cloud worlds, as an edge-cloud AI continuum. In between the edge and cloud computing worlds, an intermediate layer acts as a transition between them, as a fog layer. The benefit and value of the edge-cloud AI continuum comes from enabling the low-level sensor components, i.e., the technology devices, to realise Internet; this ability is what moves data from endpoint devices through the IoT pipeline to central servers. This ensures that data sent from endpoint devices, such as sensors, is received and understood by the next and subsequent steps in the connected environment, whether the next step for that data is to another endpoint device or a gateway or an application. In communication networks, smart switches adapt network bandwidth and quality of service to the channel communication requirements. At the core of network switches, smart schedule units provide the channelcontrol decisions according to the bandwidth and quality of service for the network traffic. In between the edge and cloud computing worlds is a transitional intermediate layer, or fog layer. The AI-supported transmission schemes can make the edge-cloud continuum more reliable and capable of working also in harsh environments (e.g., water).


Trends and challenges 

H2TRAIN includes new biosensors that are at the cutting edge of the technology development, as is the case of the cortisol, lactate and C-reactive protein. They are based on graphene, a 2D material that is functionalised for detecting biomarkers. In cross-sectional technologies, edge AI is a new trend and a dramatically growing area, but no standard products exist in the market. Edge-cloud AI continuum is a novel concept, and the approach that we follow will generate intellectual properties.


H2TRAIN holds a number of future challenges. In the area of sensing of human biomarkers, the research and innovation field is still open to new contributions for sweat tattoos with improved sensitivity and reliability. Flexible substrates for wearable sweat sensor and micro fluidic devices will be integrated on the same substrate, but the low secretion rate and rapid evaporation of sweat at rest limit the volume available to be collected in a sensor. In terms of system in package and integration in textile, i.e., textile-based platforms, this is a growing field and future challenges will focus on the screen printing process. However, in textile-based platform for bio-sensing the challenge is open to form thread-based sensors. There is a short-term need to reinforce learning in the area of embedded intelligence and, in the longer term, collaborative edge AI ecosystems will be needed.


“This is only the first project of a long series; you can be sure of this,” says Marco. The market of wearable technologies for sport and healthcare is expected to achieve doubledigit growth at least until 2030 and many opportunities (and challenges) need to be grasped. “More ideas will come, and many more partners will sit around this table.”


Download ISSUE 7 of INSIDE Magazine via this link: https://www.inside-association.eu/publications






コメント


bottom of page