top of page



SWORKS | Medical Wearables

System Description 

SWORKS offers a cloud-based solution offering Diagnosis as a Service in tandem with an innovative 12-lead wearable monitoring ECG device, and a mobile app to facilitate the remote monitoring of patients suffering from heart disease. The SWORKS system is composed of three main components, namely: 

  1. The wearable device is attached to the patient's body and its data is transmitted to the patient's smartphone that acts as an IoT gateway device.

  2. The Mobile IoT gateway application is installed on the user’s (patient’s) mobile device (smartphone or tablet) and is wirelessly connected to the wearable device in order to acquire the traces.

  3. The SWORKS Cloud services are the central orchestrator for all the services that the SWORKS platform provides. Validation Metrics 


typical electrodes configuration.jpg


ELEGANT Improvements  

ELEGANT will improve on the SWORKS use case by:

  1. Increasing code reusability of the various data operators implemented for the analysis of the ECG signal in order to extend its product line for the on-the-spot analysis of different Heart Diseases such as complex arrhythmias, acute myocardial ischaemia, acute coronary syndrome, pulmonary edema, and pharmacologically induced prolonged QT intervals. 

  2. Boosting energy efficiency and thus prolonging the battery life of the medical wearable devices since ELEGANT will enable better resource management between the IoT and Big Data analytics sides. 

  3. Enabling efficient code verification and reducing the costs of validating the source code which will be deployed on the IoT devices. 

  4. Greatly improving security and data integrity due to the decrease in data transfer from the IoT/edge to the Big Data analytics side. In addition, further security enhancements will be achieved by the newly introduced abstraction layer of ELEGANT, which will allow auxiliary authentication mechanisms to be seamlessly deployed.

This will enable faster time-to-market of medical-related code running on wearable devices; a currently time-consuming process due to restrictions enforced by the strict legislation for sensitive data management. 

the PQRST wave.png




UBI | Video Surveillance 

System Description 

Our Smart Surveillance solution involves video cameras that capture video and audio streams, collect them on a distributed message broker and then forward them for video and audio analysis. We implement security surveillance with special focus on two particular categories of surveillance equipment: a) static surveillance cameras that are connected to the network and the power distribution system of the building; these are responsible for monitoring activities in the inside and in the perimeter of buildings, and b) cameras with some kind of mobility (e.g. flying drones that capture video in the outside space of buildings, or mobile phones). The combination of these two different categories of devices can provide broader visual and audio input to be used for the analysis. A first level of processing is performed with lightweight video and audio analytics, such as face detection or a gunshot detection. A second level of processing implies advanced analytics, with the usage of ML techniques, for face recognition and number plate recognition. The main modules of the solution are the following:

  • Camera sensors and microphones: The video and audio streams for the surveillance can be created by IP cameras, cameras on embedded systems or mobile phones, drones and UAVs. For the purpose of our use case the surveillance equipment will be provided by UBI and will include a setup of 10 IP Cameras and a drone, while we will also utilize mobile phones to evaluate the optimized implementation of an embedded Java Virtual Machine that will be provided by ELEGANT.

  • Distributed Broker: Apache Kafka based broker that is used to collect data from multiple input sources and make them available for processing at the compute nodes.

  • Audio Analytics module: This module is dedicated to the processing of audio streams.

  • Video Transcoding module:  This module is responsible for making transcoding of the video in order to reduce the size or change the video type, prior to video analysis or storage.

  • Video Analytics module: This module executes video analysis, with support for face detection, facial recognition and number plate recognition. The OpenCV open source computer vision and machine learning software library is used for most advanced analytics.

ELEGANT Improvements  

ELEGANT will improve on the UBI’s Video Surveillance use case by:

  1. Enabling distributed processing for the security analytics features, thereby improving the resilience and reliability of the solution.

  2. Providing scalability to the video surveillance solution when the number of cameras increases.

  3. Ability to utilize edge resources, as ELEGANT will enable the execution of the same unmodified source code on all devices.

  4. Minimizing latency by utilizing edge resources.

  5. Exploiting edge resources for in-situ processing for privacy reasons.

  6. Minimizing cost with the usage of edge resources of various types and architectures.



UND/CNIT | Large-scale secure smart metering

System Description 

UND is managing a wide coverage IoT network in Italy, based on LoRaWAN (Long Range Wide Area Network) technology. This technology includes a very robust and energy-efficient modulation scheme, which operates in the Industrial Scientific and Medical (ISM) bands and can connect thousands of IoT devices, called End-Devices (EDs), with a cover range of several kilometres. The communication starts with EDs sending packets to Gateways (GWs) which, in turn, forward them to the Network Server (NS). NS is responsible for processing packets, and forwards information data to the IoT applications server and to the Network Controller (NC).



Some examples of application services provided by this network are: i) water metering, ii) energy consumption metering, iii) GPS tracking, etc. Currently, the total number of connected smart objects is 1862, while the total number of gateways is 138. The network also collects raw packets from not connected devices (devices of different operators), with a total amount of 89.528, monitored devices generating about 400 million packets per year. The network server is located in Rome, together with the NC, developed by UND, which is part of the UniOrchestra suite, it also supporting the management of the network. Note that the NC is somehow equivalent to a special application server responsible of presenting statistical information about the network operation to the operator.

We plan to extend the LoRaWAN architecture in order to support code motion among EDs, GWs and server applications, by means of the ELEGANT framework and API for IoT and Big Data. In particular, with reference to our large-scale smart metering use case, the following three functionalities will be addressed:

  1. A self-healing LoRaWAN network architecture in order to provide resilience under critical situations such as earthquakes, fires, etc. Our approach tries to provide a case-generic framework valuable for a wide set of use cases with no need to bloat the network with additional devices. The idea is exploiting these functionalities when the NS unavailable and reduce the computational load of the central server.

  2. ED profiling and environmental interference prediction for improving resource allocation. Network performance in LoRaWAN networks are critically affected by resource allocations, in terms of SFs and operation channels, as a function of the per-node link quality and interference conditions of the network. The objective here is to transform the LoRaWAN architecture into a holistic Big Data/IoT framework that will automate resource provisioning and dynamic edge-processing management, by also offering programmable interfaces to operators for a rapid prototyping and experimentation of resource allocation schemes.

  3. Security applications for metering. We will deploy distributed monitoring functions for early detection of security threats. Moreover, we will consider the possibility of deploying local application servers in a given geographical area for performing data aggregation functionalities.


ELEGANT Improvements  

We will demonstrate how IoT/Big Data applications can be deployed as a single system, by uniformly programming and orchestrating code’s execution, based on the performance, energy, security and optimization space.

ELEGANT will improve the large-scale secure smart metering use case by:

  1. Increasing code reusability, decomposing NS, NC and Application server functionalities by exploiting the instantiation of these functionalities at different levels.

  2. Defining a network operating system able to dynamically allocate network resources where needed and to reconfigure in case of failure or congestion. Apply ML techniques, and optimum SF allocation, to optimize network.

  3. Greatly improving security and data integrity due to the reduction of the data transfers from the IoT/edge to the Big Data analytics side.



KTM Innovation | Secure Smart Riding

System Description 

The figure provides a high-level description of KTM’s smart two-wheelers along with their interactions with various peripherals.



As shown, the motorbike on-board system collects and processes data from miscellaneous sensors through the CAN bus, mobile devices, and other peripherals. Similarly, different parts of the processing pipeline may execute on the motorbike, the mobile device, and to the back-end analytics servers. Since the on-board devices can have direct internet access through WiFi or via the mobile devices, it provides the opportunity to remotely control the data operators (in the form of executed code) running on the edge or at the back-end analytics. This is of great importance since it will dramatically assist in controlling the data communications between different entities in the connected ecosystem as well as achieving higher levels of performance when needed. Below is a small number of different scenarios where the ability to control the executed code can improve the overall performance:

  1. Stress-testing of motorbikes: During stress-testing the bikes are equipped with extra sensors that produce vast amounts of data and require extreme compute power to be processed at real-time.

  2. Motorsport bike tuning: Prior to races, motorbikes are equipped with special sensors which provide information for tuning during the race. Similar to the first scenario these sensors produce vast amounts of data and require extreme processing.

  3. Processing of radar data in real-time and adaptive trace generation and processing for diagnostics.

ELEGANT Improvements  

ELEGANT will improve on KTM’s Secure Smart Riding use case by:

  1. Protecting its IoT connected ecosystem against cyber-attacks to ensure the dependable execution of its fleet of vehicles.

  2. Increasing the performance and energy efficiency of its data analytics through dynamic code motion.

  3. Advancing the reliability and dependability of its connected ecosystem through ELEGANT’s global orchestration.

bottom of page