Main ongoing research projects

MAIN ONGOING RESEARCH PROJECTS

The research activity currently carried by Prof. Enzo Baccarelli is corroborated by the involvement in five main ongoing research projects, namely,

• the V-Fog project

• the V-Fog2 project

• the GAUChO project

• the SoFT project

• the DeepFog project

The V-Fog and V-Fog2 projects

Delivering cloud-supported real-time Big Data streaming services to Vehicular Clients (VCs) must cope with delay and delay-jitter issues, while mining large volume of data. Fog computing is a (still unexplored) emerging paradigm, which aims at distributing small-size self-powered networked data centers (e.g., Fog nodes) between resource-rich remote Clouds and resource-limited smartphone-equipped VCs, in order to perform context/content aware energy-efficient data mining and dissemination. Motivated by this consideration, the “Vehicular Fog for energy-efficient QoS mining and dissemination of multimedia Big Data streams” (V-Fog) and its follow-up V-Fog2 are two three-year projects funded by Sapienza University of Rome. They aim to define, design and validate integrated resource-management and data-mining distributed self-adaptive algorithms for Fog-supported vehicular networks. The final goal is the energy-efficient support of real-time Big-Data-streaming-type Future Internet Applications, such as multimedia human activity recognition and infotainment interactive services (e.g., VTube). In this regard, the actual (still unexplored) transport capabilities promised by the novel paradigm of multipath-TCP is investigated and a cognitive approach is pursued for the integrated design of the V-Fog architecture.

A sketch of the reference scenario of the V-Fog and V-Fog2 projects

Overall, the main goals of the V-Fog and V-Fog2 projects are (see the reference figure):

• • design and validate novel distributed and adaptive real-time algorithms for the QoS mining of environmental multimedia Big Data streams. The developed algorithms must be capable to quickly self-scale up/down;

  • design and validate minimum-bandwidth algorithms for the QoS energy-efficient migration of Virtual Machines to/from Fog nodes over multipath-TCP 5G connections;

  • design and validate adaptive Machine Learning-based mining algorithms for the real-time context-aware detection and classification of sensor-acquired human activities of specific social interest.

The PRIN2017 GAUChO project and the related virtFogSim simulator by the BroadComLab

Fog Computing (FC) is an emerging paradigm that extends Cloud Computing towards the edge of the network. Specifically, FC refers to a distributed computing infrastructure confined on a limited geographical area in which some applications/services run directly at the network edge in smart end-devices. The goal of FC is to improve efficiency and reduce the amount of data that needs to be transported to the Cloud for massive data processing, analysis and storage. However, in latency-sensitive and energy-efficient (i.e., green) applications, the FC paradigm per se might not be appropriate and (at least) part of the computation has to be transferred to the sensor/actuator end-devices level. Here, decisions must be usually taken in a very short time and energy and power constraints constitute a limiting factor. Furthermore, the design of efficient solutions within FC also requires investigate a novel communication/networking paradigm, called Fog Networking (FN), in order to meet specific configurability, adaptability, flexibility and energy/spectrum-efficiency constraints. Being currently FC and FN features designed, optimised and implemented independently each other, the GAUChO project (web page: https://www.gaucho.unifi.it) is a three-year project funded by the Italian MIUR under the PRIN2017 umbrella.

The logo of the PRIN2017 GAUChO project

GAUChO aims at designing a novel distributed and heterogeneous architecture able to functionally integrate and jointly optimise FC and FN capabilities in the same platform. The joint FC-plus-FN architecture, representing the overall outcome of the project, aims at supporting low-latency and energy-efficiency as well as security, self-adaptation. and spectrum efficiency by means of a strict collaboration among end-devices and FC-plus-FN units in a same integrated platform. In addition, the development of suitable analytic methods and definition of appropriate techniques will enable extra relevant characteristics of the FC-plus-FN platform including ubiquity, decentralised management, cooperation, proximity to end users, dense geographical distribution, efficient support for mobility and real-time applications. To achieve this goal, the GAUChO project foresees to address several relevant and challenging research topics that require skills and knowledge in different scientific fields. For this reason, four Research Units characterised by different but complementary scientific expertise are involved on it.

Overall, the scientific contributions of the GAUChO project cover the following main aspects (see the following figure):

• efficient schemes for the coordinated management of resources and interference in heterogeneous wireless communication systems;

• joint optimisation of communication and computing capabilities to support energy-efficient management and self-reconfigurations;

• a learning modality permitting software agents to detect variations within the integrated FC-plus-FN platform;

• model-free fault diagnosis systems and comprehensive methodology integrating intelligence-based mechanisms for optimally managing energy consumption, detecting changes in environment system under inspection, and evaluating and mitigating the possible occurrence of faults affecting the end-devices computing units.

A sketch of the reference scenario of the PRIN2017 GAUChO project simulated by the virtFogSim tool

The GAUChO project is expected to have a significant technological impact on many up-to-date and relevant families of technologies, such as Smart Wireless Sensor Networks, Smart Objects of Internet-of-Things and emerging Industry 4.0 technological platforms. The project will allow the FC-plus-FN paradigm to enter into a new phase where real-world problems emerging from complex applications are addressed and effectively solved. It is expected that project outcomes will convert from basic research into mass production during the years 2018-2025 providing significant economic benefits to masses of potential end users, together with the added value of the offered application services, limited cost of the communication and processing infrastructure.

A software simulation package (namely, the virtFogSim tool) of the GAUChO technological platform is under development by the research team of Prof. Enzo Baccarelli at the BroadComLab of the DIET department. The virtFogSim package and the related user guide may be downloaded for free by accessing the section Downloadable Packages of this web page.

The SoFT project

Up to date, Social Internet of Things (SIoT) and Fog Computing (FC) are two standing-alone technological paradigms under the realm of the Future Internet. SIoT relies on the self-establishment and self-management of inter-thing social relationships, in order to guarantee scalability to large IoT networks composed by both human and non-human agents. FC extends the Cloud capabilities to the access network, in order to allow resource-poor IoT devices to support delay-sensitive applications. Motivated by these considerations, Sapienza University of Rome recently funded the two-year project: “Fog of Social IoT (SoFT)” . The project aims at proposing their integration into the novel SoFT platform. Specifically, FC supports natively three main services, namely thing virtualisation, Thing-to-Fog task offloading, and inter-Fog resource pooling. In principle, these services could be efficiently exploited, in order to implement the SIoT social network as an overlay network of thing clones, that entirely relies on the bandwidth/computing resources of the supporting Fog Nodes (FNs). So doing, the native resources of the physical things could be employed only for the synchronising the corresponding Fog-hosted clones. This is, indeed, the main idea behind the proposed SoFT paradigm.

The reference scenario of the SoFT project

Overall, main goals of the SoFT project are expected to be to (see the reference figure):

• formalise the main building blocks and functionalities of the proposed SoFT technological platform. It merges the physical things at the IoT layer and their virtual clones at the Fog layer into a cyber-physical overlay network of social clones;

• design and validate through software simulations the performance of a small-scale SoFT prototype, and compare its energy-vs.-delay performance with the corresponding ones of the state-of-the-art;

• design and validate through software simulations novel distributed machine learning and deep learning algorithms for the analysis and forecasting of Big Data diffusion through the SoFT social network.

The DeepFog project - Optimized distributed implementation of Deep Learning models over networked multitier Fog platforms for IoT stream Applications

The incoming IoT big data era requires efficient and resource-saving mining of large sets of distributed data. To this end, a viable approach could be to attempt to resort to the synergic exploitation of two emerging paradigms, namely, Deep Neural Networks (DNNs) with early exits and Fog Computing. From an IoT perspective, the most appealing feature of DNNs is its hierarchical layer-wise organization that is exploited for feature learning and/or pattern analysis/classification, potentially matching the hierarchical organization of the underlying Fog-supported execution platform. However, both the training and inference phases of current DNNs are too cumbersome to be executed by IoT devices, and their remote and centralized implementation induces large communication delays and is bandwidth-wasting. Motivated by these considerations, the two-year DeepFog project aims at developing the integration of the emerging paradigm of the Conditional DNNs (CDNNs) with early exits with the novel Fog Computing (FC) platform that is hierarchically organized in multiple networked tiers. For this purpose, the exploitation of the powerful and recent paradigm of Federated Learning (FL) is also utilized for the distributed and asynchronous Fog-supported training of the underlying CDNN under resource and fairness constraints.

The multi-tier computing networked architecture envisioned for proposed DeepFog technological platform. The

platform aims at integrating cloud, fog, edge (5G wireless access points), and IoT devices over a distributed realm.

Overall, main goals of the DeepFog project are to:

• formalize the main building blocks and functionalities of the proposed DeepFog technological platform, by designing the most suitable algorithms for the CDNN training;

• design and validate through software simulations a novel asynchronous fast-adaptive FL protocol for the training of CDNNs for Fog supported IoT stream applications;

• design and validate through extensive numerical tests the performance of the overall resulting optimized DeepFog technological platform, and compare its energy-vs.-complexity-vs.-accuracy performance under some real-world datasets for IoT-oriented real-time big-data applications.