Periodic Reporting for period 1 - IntellIoT (Intelligent, distributed, human-centered and trustworthy IoT environments)
Berichtszeitraum: 2020-10-01 bis 2022-03-31
The Internet of Things (IoT), as a cornerstone of digitalization and data-driven business models, has successfully spawned cloud-centric approaches where IoT cloud platforms are central points of data collection and processing.
This cloud-centric IoT model however has limitations:
(i) unreliable cloud connectivity impedes dependable end-to-end applications,
(ii) limited bandwidth restricts the amount of data that can be processed,
(iii) high round-trip times prevent real-time operation,
(iv) high cost of data transport and intake, and
(v) privacy and trust concerns.
IntellIoT focuses on: Agriculture, where farming machinery (e.g. tractors) is semi-autonomously operated in conjunction with supporting devices. Healthcare, where patients are monitored by sensors to receive instantaneous healthcare advice and interventions from virtual advisors. Manufacturing, where plants are highly automated and shared by multiple tenants who utilize machinery from third-party vendors.
In all three use case areas, a human expert plays a key role in controlling and teaching the AI-enabled autonomous systems, for instance since clinicians need to monitor plausibility of the AI, when autonomous agricultural vehicles encounter unknown obstacles, and when manufacturing systems deal with novel production contexts.
Objective 1: Create a self-aware and semi-autonomous multi-agent system over an optimized computation and communication infrastructure that manages compositions of IoT/edge devices in closed-loop with the network.
Objective 2: Enable ultra-reliable low-latency communication over heterogeneous networks to enable tactile (real-time) and contextual (adaptive) interaction between IoT devices, humans, and services.
Objective 3: Enable semi-autonomous IoT applications by leveraging distributed AI algorithms under compute, storage, mobility and bandwidth constraints and by integrating the human-in-the-loop for safety, assistance and continuous improvement of AI.
Objective 4: Enable security, privacy and trust by design with continuous assurance monitoring, assessment and certification as an integral part of the system, providing trustworthy integration of third party IoT devices and services.
Objective 5: Development of a reference implementation of the IntellIoT framework, demonstrated and evaluated in the three use case areas: agriculture, healthcare and manufacturing.
Objective 6: Promote and exploit the IntellIoT framework through contribution to standards and open source as well as by building an active IoT ecosystem supported by two Open Calls and focused dissemination and exploitation activities.
This cloud-centric IoT model however has limitations:
(i) unreliable cloud connectivity impedes dependable end-to-end applications,
(ii) limited bandwidth restricts the amount of data that can be processed,
(iii) high round-trip times prevent real-time operation,
(iv) high cost of data transport and intake, and
(v) privacy and trust concerns.
IntellIoT focuses on: Agriculture, where farming machinery (e.g. tractors) is semi-autonomously operated in conjunction with supporting devices. Healthcare, where patients are monitored by sensors to receive instantaneous healthcare advice and interventions from virtual advisors. Manufacturing, where plants are highly automated and shared by multiple tenants who utilize machinery from third-party vendors.
In all three use case areas, a human expert plays a key role in controlling and teaching the AI-enabled autonomous systems, for instance since clinicians need to monitor plausibility of the AI, when autonomous agricultural vehicles encounter unknown obstacles, and when manufacturing systems deal with novel production contexts.
Objective 1: Create a self-aware and semi-autonomous multi-agent system over an optimized computation and communication infrastructure that manages compositions of IoT/edge devices in closed-loop with the network.
Objective 2: Enable ultra-reliable low-latency communication over heterogeneous networks to enable tactile (real-time) and contextual (adaptive) interaction between IoT devices, humans, and services.
Objective 3: Enable semi-autonomous IoT applications by leveraging distributed AI algorithms under compute, storage, mobility and bandwidth constraints and by integrating the human-in-the-loop for safety, assistance and continuous improvement of AI.
Objective 4: Enable security, privacy and trust by design with continuous assurance monitoring, assessment and certification as an integral part of the system, providing trustworthy integration of third party IoT devices and services.
Objective 5: Development of a reference implementation of the IntellIoT framework, demonstrated and evaluated in the three use case areas: agriculture, healthcare and manufacturing.
Objective 6: Promote and exploit the IntellIoT framework through contribution to standards and open source as well as by building an active IoT ecosystem supported by two Open Calls and focused dissemination and exploitation activities.
In the covered period, the work performed and the resulting achievements are grouped per objective as follows:
Obj. 1:
- Formulated our vision of an integrated framework for Intelligent IoT Environments,
- Developed Hypermedia Multi-Agent Systems (HyperMAS) infrastructure that supports the registration of artifacts via W3C WoT TD, their discovery, end-user programming, and execution
- Integration of HyperMAS in all 3 use cases
- Implementation and integration of edge infrastructure and orchestration
Obj. 2:
- Deployed OpenAirInterface 5G infrastructure and enhanced towards ultra-reliable and low latency communication (URLLC)
- Implemented 5G network slice management in Communication Resource Manager
- Developed REST interfaces and containerization for TSN controller as preparation for integration with 5G
Obj. 3:
- Developed resource-aware client scheduling algorithm for federated learning
- Developed energy-efficient model compression & splitting for collaborative inference
- Developed algorithm for adaptively optimizing split learning parameters to minimize total energy depending on the channel conditions
- Algorithm improved for increased accuracy on Holo-Stylus. Therefore, stabilized the Bluetooth connection between HoloLens and Holo-Stylus. We achieved an accuracy between 6-10 mm so far.
Obj. 4:
- Developed & integrated Security Assurance Platform (SAP) in IntellIoT
- Defined an architecture for DLT integration and analyzed different DLT platform options
- Designed of communication protocols for integrating DLT in resource-limited IoT devices, by studying tradeoff between trust and latency
- Implemented & integrated Moving Target Defense system (auto changing network configurations as security measure)
Obj. 5:
- Use cases were defined through scenario scenes to enable validation of all aspects of framework
- Requirements were derived to create a basis for evaluation criteria
- Integration of IntellIoT framework components and use case specific hardware
Obj. 6:
- Open Call 1 extensive communication and promotion campaign (204 applications) supported by Guide for Applicants
- Evaluation & selection of applications and setup of implementation phase
- Dissemination and communication targets as pre-defined are well under way
- Conducted multiple requirements and exploitation workshops with end-user groups
- Internal exploitation (e.g. Siemens acquired ~400 k € in research-to-production projects)
Obj. 1:
- Formulated our vision of an integrated framework for Intelligent IoT Environments,
- Developed Hypermedia Multi-Agent Systems (HyperMAS) infrastructure that supports the registration of artifacts via W3C WoT TD, their discovery, end-user programming, and execution
- Integration of HyperMAS in all 3 use cases
- Implementation and integration of edge infrastructure and orchestration
Obj. 2:
- Deployed OpenAirInterface 5G infrastructure and enhanced towards ultra-reliable and low latency communication (URLLC)
- Implemented 5G network slice management in Communication Resource Manager
- Developed REST interfaces and containerization for TSN controller as preparation for integration with 5G
Obj. 3:
- Developed resource-aware client scheduling algorithm for federated learning
- Developed energy-efficient model compression & splitting for collaborative inference
- Developed algorithm for adaptively optimizing split learning parameters to minimize total energy depending on the channel conditions
- Algorithm improved for increased accuracy on Holo-Stylus. Therefore, stabilized the Bluetooth connection between HoloLens and Holo-Stylus. We achieved an accuracy between 6-10 mm so far.
Obj. 4:
- Developed & integrated Security Assurance Platform (SAP) in IntellIoT
- Defined an architecture for DLT integration and analyzed different DLT platform options
- Designed of communication protocols for integrating DLT in resource-limited IoT devices, by studying tradeoff between trust and latency
- Implemented & integrated Moving Target Defense system (auto changing network configurations as security measure)
Obj. 5:
- Use cases were defined through scenario scenes to enable validation of all aspects of framework
- Requirements were derived to create a basis for evaluation criteria
- Integration of IntellIoT framework components and use case specific hardware
Obj. 6:
- Open Call 1 extensive communication and promotion campaign (204 applications) supported by Guide for Applicants
- Evaluation & selection of applications and setup of implementation phase
- Dissemination and communication targets as pre-defined are well under way
- Conducted multiple requirements and exploitation workshops with end-user groups
- Internal exploitation (e.g. Siemens acquired ~400 k € in research-to-production projects)
The progress beyond the state of the art is as follows:
Obj 1.:
- Combination of state-of-the art multi-agent systems with REST architectural style, t
- Developed the basis for advanced, dynamic resource management within private edge environments.
Obj. 2:
- Enhanced the OpenAirInterface 5G component, as a central component of the IntellIoT framework, toward tactile IoT support, by improving its ultra-low latency and reliability communication.
- Developed the basis of the Communication Resource Manager that will realize a novel data-driven approach
Obj. 3:
- We developed multiple algorithms that enable scalable machine learning techniques, which support application-specific target accuracies
- Development of algorithms that account for the wireless resources availability, on-device energy, storage and computing restrictions as a co-design of machine learning.
- Facilitating the integration of the human-in-the-loop through advancing the AR/VR stylus accuracy.
Obj. 4:
- Development of novel communication protocols for DLT-based IoT environments.
- Developed the basis for novel moving target defense strategies that re-configure the network of the system to prevent security vulnerabilities and attacks.
Obj. 5:
- Integration of the novel IntellIoT framework that combines various next generation IoT technologies and its deployment in 3 heterogeneous use case setups
Potential impacts:
- The developed concept of the human-in-the-loop in all 3 use cases and throughput the IntellIoT framework will improve user acceptance of the IoT environments and combined with advanced security and DLT solutions will protect users.
- The novel HyperMAS architecture for next generation IoT environments will be contributed to W3C WoT standards and combined with the communication & computation infrastructure based on 5G and edge computing will facilitate the development of semi-autonomous IoT applications.
- The excellent communication and outreach activities (e.g. Medium channel) driven by IntellIoT (leading to > 200 applications in the first Open Call) demonstrate the beginnings of a relevant ecosystem building that will attract key IoT players and allow for novel business models.
Obj 1.:
- Combination of state-of-the art multi-agent systems with REST architectural style, t
- Developed the basis for advanced, dynamic resource management within private edge environments.
Obj. 2:
- Enhanced the OpenAirInterface 5G component, as a central component of the IntellIoT framework, toward tactile IoT support, by improving its ultra-low latency and reliability communication.
- Developed the basis of the Communication Resource Manager that will realize a novel data-driven approach
Obj. 3:
- We developed multiple algorithms that enable scalable machine learning techniques, which support application-specific target accuracies
- Development of algorithms that account for the wireless resources availability, on-device energy, storage and computing restrictions as a co-design of machine learning.
- Facilitating the integration of the human-in-the-loop through advancing the AR/VR stylus accuracy.
Obj. 4:
- Development of novel communication protocols for DLT-based IoT environments.
- Developed the basis for novel moving target defense strategies that re-configure the network of the system to prevent security vulnerabilities and attacks.
Obj. 5:
- Integration of the novel IntellIoT framework that combines various next generation IoT technologies and its deployment in 3 heterogeneous use case setups
Potential impacts:
- The developed concept of the human-in-the-loop in all 3 use cases and throughput the IntellIoT framework will improve user acceptance of the IoT environments and combined with advanced security and DLT solutions will protect users.
- The novel HyperMAS architecture for next generation IoT environments will be contributed to W3C WoT standards and combined with the communication & computation infrastructure based on 5G and edge computing will facilitate the development of semi-autonomous IoT applications.
- The excellent communication and outreach activities (e.g. Medium channel) driven by IntellIoT (leading to > 200 applications in the first Open Call) demonstrate the beginnings of a relevant ecosystem building that will attract key IoT players and allow for novel business models.