Periodic Reporting for period 2 - A-WEAR (A network for dynamic WEarable Applications with pRivacy constraints)
Periodo di rendicontazione: 2021-01-01 al 2023-06-30
Rapid technology advancement, economic growth, industrialization, and evolution in the information technology have paved the way for developing a new niche of small body-worn personal devices gathered together under the title of wearable technologies. The growing interest from end-users has resulted in introducing the first generation of mass consumer wearables. Evidently, the trailblazing ones were not designed with strict restriction (energy efficiency, privacy, security, communication, etc.) in mind, and, thus, research on this topic has remained fragmented. Advanced and sophisticated batteries, lightweight security enablers, and communication technologies could be already procurable on the wearable devices. Additional solutions for efficient utilization of the processing power, especially for recent computationally and power-hungry applications, are still a white spot on the wearable technology roadmap. A-WEAR EU project has enhanced the understanding of how the superimposition of those technologies would improve wearable devices' performance metrics, including energy efficiency, and has explored novel and emerging research solutions such as edge computing's paradigm and collaborative positioning.
In addition to the scientific breakthroughs (detailed below), A-WEAR project has also educated 15 Early Stage Researchers, with most of them having already graduated with a PhD degree and the rest expected to graduate during this Fall or early next year. Dissemination activities targeted not only the scientific community, but also the general public, including the end-users in wearable domain, through multi-dimensional social media activities such as A-WEAR Youtube and Twitter channels and blog entries at the project webpage.
The project's main two target groups in wearables applications were the social applications, such as eHealth and social networking, and the industrial applications, such as automation halls, industrial robotics, and the automotive industry. Such applications were chosen because they had different requirements in terms of computational costs, communications latency, precision, security, and privacy of both communication and localization and related tradeoffs.
As of 15.8.2023 8 out of the 15 A-WEAR ESRs (i.e. 53%) have already defended their double PhD degree thesis and published their PhD theses in open access. Two more ESRs have already submitted their PhD manuscript for review, and three more are expected submit their thesis manuscript by November 2023; the last one is expected to have his thesis submitted to review by early 2024.
The A-WEAR scientific objectives have been as follows:
O1. To provide the society with general knowledge of dynamic wearable networks in terms of localization, connectivity, privacy, and security;
O2. To identify vulnerabilities and offer innovative solutions in crowdsourced-, cloud-, edge-, and fog-based wearable architectures in the telecommunications area;
O3. To design and develop privacy-enhanced and location-aware wearable technologies;
O4. To improve the communications between wearable devices, their smartphone gateways and the infrastructure networks;
O5. To develop new open-source software for wearables in social/eHealth/industrial applications.
In terms of scientific breakthroughs, we would like to pinpoint the following strong results within A-WEAR network: proposing novel lossy compression methods for performance-restricted devices, analyzing deep learning-based new localization and optimization methods, highly improving the performance metrics of indoor collaborative and standalone wireless positioning methods, deeply investigating the tradeoffs between privacy and utility and between performance and energy consumption in a variety of wearable scenarios and with various technologies(UWB, BLE, WiFi, LoRa), analyzing the mobility-awareness in directional deafness within mmWave communications and modeling system-level dynamics of direct Extended Reality sessions over mmWave links with wearable helmets and glasses, developing machine-learning-based algorithms and techniques for detecting hand tremors and bradykinesia in Parkinson's patients and for analyzing electroencephalogram data, which is used to study brain activity using wearable devices and implementing a first A-WEAR bracelet embedding machine learning algorithms for e-health applications, investigating the integration of artificial intelligence to facilitate seamless communication between terrestrial (e.g. wearables) and non-terrestrial networks (e.g. UAV-based and satellite-based), exploring cryptographic protocols to enhance the communication privacy and security using wearable devices and constrained devices, optimizing the placement of social digital twins in edge internet-of-things and wearable networks, and proposing novel unsupervised learning approaches for device-to-device-assisted multicast scheduling in wearable networks and beyond.
In addition to the scientific breakthroughs (detailed below), A-WEAR project has also educated 15 Early Stage Researchers, with most of them having already graduated with a PhD degree and the rest expected to graduate during this Fall or early next year. Dissemination activities targeted not only the scientific community, but also the general public, including the end-users in wearable domain, through multi-dimensional social media activities such as A-WEAR Youtube and Twitter channels and blog entries at the project webpage.
The project's main two target groups in wearables applications were the social applications, such as eHealth and social networking, and the industrial applications, such as automation halls, industrial robotics, and the automotive industry. Such applications were chosen because they had different requirements in terms of computational costs, communications latency, precision, security, and privacy of both communication and localization and related tradeoffs.
As of 15.8.2023 8 out of the 15 A-WEAR ESRs (i.e. 53%) have already defended their double PhD degree thesis and published their PhD theses in open access. Two more ESRs have already submitted their PhD manuscript for review, and three more are expected submit their thesis manuscript by November 2023; the last one is expected to have his thesis submitted to review by early 2024.
The A-WEAR scientific objectives have been as follows:
O1. To provide the society with general knowledge of dynamic wearable networks in terms of localization, connectivity, privacy, and security;
O2. To identify vulnerabilities and offer innovative solutions in crowdsourced-, cloud-, edge-, and fog-based wearable architectures in the telecommunications area;
O3. To design and develop privacy-enhanced and location-aware wearable technologies;
O4. To improve the communications between wearable devices, their smartphone gateways and the infrastructure networks;
O5. To develop new open-source software for wearables in social/eHealth/industrial applications.
In terms of scientific breakthroughs, we would like to pinpoint the following strong results within A-WEAR network: proposing novel lossy compression methods for performance-restricted devices, analyzing deep learning-based new localization and optimization methods, highly improving the performance metrics of indoor collaborative and standalone wireless positioning methods, deeply investigating the tradeoffs between privacy and utility and between performance and energy consumption in a variety of wearable scenarios and with various technologies(UWB, BLE, WiFi, LoRa), analyzing the mobility-awareness in directional deafness within mmWave communications and modeling system-level dynamics of direct Extended Reality sessions over mmWave links with wearable helmets and glasses, developing machine-learning-based algorithms and techniques for detecting hand tremors and bradykinesia in Parkinson's patients and for analyzing electroencephalogram data, which is used to study brain activity using wearable devices and implementing a first A-WEAR bracelet embedding machine learning algorithms for e-health applications, investigating the integration of artificial intelligence to facilitate seamless communication between terrestrial (e.g. wearables) and non-terrestrial networks (e.g. UAV-based and satellite-based), exploring cryptographic protocols to enhance the communication privacy and security using wearable devices and constrained devices, optimizing the placement of social digital twins in edge internet-of-things and wearable networks, and proposing novel unsupervised learning approaches for device-to-device-assisted multicast scheduling in wearable networks and beyond.
A-WEAR found and recruited 15 ESRs, with 8 of them having already defended their PhD (under double-degree programmes) and the rest expected to complete their PhD by early next year. Also 133 publications (scientific journal and conference publications as well as open-access software and datasets) have been accepted and publishedin various peer-reviewed international forums; several ones are still under submission. All published articles and datasets are also available in open-access under A-WEAR Zenodo channel. The research topics addressed in A-WEAR include the following main directions (the related objectives are shown in brackets):
Develop machine learning solutions for wearable computing (O1, O4, O5)
Evaluate the wearable data impact on future wireless networks and use of wearables for public safety scenarios (O1, O2, O5)
Various applications of wearables for COVID-19 localization and detection; (O3, O5)
Test modern information security for the use on wearable devices; (O3, O4)
Provide new datasets for indoor positioning studies as well as new compression methods; (O2)
Look into the opportunities of wearable devices for crowdsourced data collection; (O2,O4)
Study tradeoffs of privacy and location accuracy in opportunistic wearable networks.(O3)
Study multicast transmissions in GHz ranges (O4)
Analyze of applications of wearables to Parkinson’s disease and neurocognitive disorder (eHealth); (O5)
Develop machine learning solutions for wearable computing (O1, O4, O5)
Evaluate the wearable data impact on future wireless networks and use of wearables for public safety scenarios (O1, O2, O5)
Various applications of wearables for COVID-19 localization and detection; (O3, O5)
Test modern information security for the use on wearable devices; (O3, O4)
Provide new datasets for indoor positioning studies as well as new compression methods; (O2)
Look into the opportunities of wearable devices for crowdsourced data collection; (O2,O4)
Study tradeoffs of privacy and location accuracy in opportunistic wearable networks.(O3)
Study multicast transmissions in GHz ranges (O4)
Analyze of applications of wearables to Parkinson’s disease and neurocognitive disorder (eHealth); (O5)
The main progress beyond the state of the art in relationship with the project objectives can be summarized as follows:
A new model for the analysis of system-level dynamics of direct extended reality sessions over mmWave links (O1).
A new algorithm for optimal placement of social Digital Twins in edge IoT networks as well as new dat compression algorithms (O2);
Several improvements for indoor positioning algorithms and a testbed for ground-truth systems for anchor-based indoor localization (O3);
A deep learning-based localization and handover optimization for 5G NR networks (O4);
A novel fractional order model for exoskeleton hand control, a novel machine-learning algorithms for posture identification of Obstructive Sleep Apnea patients and a robust technique to detect COVID-19 using chest X-ray images (O5);
Our publications provide valuable insights on human and organizational factors in technological changes caused by the Internet of Wearable Things on communications, localization, and other technological aspects and pave the way towards safer products and services.
A new model for the analysis of system-level dynamics of direct extended reality sessions over mmWave links (O1).
A new algorithm for optimal placement of social Digital Twins in edge IoT networks as well as new dat compression algorithms (O2);
Several improvements for indoor positioning algorithms and a testbed for ground-truth systems for anchor-based indoor localization (O3);
A deep learning-based localization and handover optimization for 5G NR networks (O4);
A novel fractional order model for exoskeleton hand control, a novel machine-learning algorithms for posture identification of Obstructive Sleep Apnea patients and a robust technique to detect COVID-19 using chest X-ray images (O5);
Our publications provide valuable insights on human and organizational factors in technological changes caused by the Internet of Wearable Things on communications, localization, and other technological aspects and pave the way towards safer products and services.