1,721,005 research outputs found
Wardrop Equilibrium in Discrete-Time Selfish Routing with Time-Varying Bounded Delays
Full text linkThis paper presents a multi-commodity, discrete-
time, distributed and non-cooperative routing algorithm, which is
proved to converge to an equilibrium in the presence of
heterogeneous, unknown, time-varying but bounded delays.
Under mild assumptions on the latency functions which describe
the cost associated to the network paths, two algorithms are
proposed: the former assumes that each commodity relies only on
measurements of the latencies associated to its own paths; the
latter assumes that each commodity has (at least indirectly) access
to the measures of the latencies of all the network paths. Both
algorithms are proven to drive the system state to an invariant set
which approximates and contains the Wardrop equilibrium,
defined as a network state in which no traffic flow over the
network paths can improve its routing unilaterally, with the latter
achieving a better reconstruction of the Wardrop equilibrium.
Numerical simulations show the effectiveness of the proposed
approach
Control methods for safe and efficient cyber-physical systems
Full text linkThe present manuscript represents a report of the main research activities done by the candidate during the three years of his PhD cursus studiorum.
The work revolves around the concept of Cyber-Physical Systems (CPS), a class of systems in which the interaction between their digital domain, constituted by connected devices capable of computing, and a physical process plays a fundamental role in their operation. The deep linkage between the cyber and the physical parts of a CPS makes their study appealing for several research fields as Automation and Computer Science, as properties such as stability and robustness are paired with concepts as information integrity and service availability.
Due to their very broad definition, CPS are studied and applied in the most heterogeneous domains, spacing from power systems and smart factories to healthcare and autonomous vehicles. The present work explores a total of four case studies that the candidate analysed during his PhD studies, covering energy & power management system, spacecraft control and selfish routing over dynamical networks.
The typical goal of a control system designed for a CPS is the one of attaining the desired, optimal, behaviour in the most efficient way possible, while also constraining the system evolution into a region considered to be safe for the system itself and the environment around it. This simple idea is behind the development of the first work presented, which was carried out by the candidate in the scope of the research project H2020 ATENA (regarding Critical Infrastructure Protection). The work develops a control solution for the Risk-Aware and Efficient operation of the Power Distribution Network, exploiting the presence of innovative devices as Electrical Storage Systems. In this application scenario, the candidate designed an Economic Model Predictive Controller (EMPC) for the purpose of increasing the resiliency of the service provision, by automatically reconfiguring the power network in response to, predicted or ongoing, adverse events, as malicious attacks or faults. The specifics of the case study were refined during the project thanks to the continuous interaction with the Israel Electric Company (IEC), principal industrial end user of ATENA.
A more user-centric approach is taken for the development of the second controller proposed in this work, as it was designed to manage the heating and power appliances of a smart building, integrating also features as electric vehicles charging and demand side management capabilities. The methodology chosen for this second controller was still EMPC, as it offered the possibility of explicitly consider operational and logical constraints imposed by the specific case study, while also exploiting short-term prediction of the exogenous signals interacting with the system.
The third example of CPS studied by the candidate in this work was a multi-body satellite system. Thanks to a collaboration opportunity with the manufacturer Thales Alenia Space Italia, a real case study of interest for the company was analysed, leading to the development of a control scheme for a life-support system that could connect to orbiting satellites to extend their operative life and upgrade their capabilities. The control scheme developed is based on feedback linearization, under which it was proven that the two interconnected spacecraft may operate in parallel without requiring communication or information exchanges. The life-support is in fact able to remove its effects on the original satellite dynamics by applying a compensating control action that reconstructs the original system behaviour.
The fourth, and final, study presented was completed in the scope of the H2020 EU-Korea Project 5G-ALLSTAR (regarding the integration of satellite communications and 5G) and deals with the problem of selfish routing and load balancing in heterogeneous networks, to enable 5G Multi-Connectivity. The network studied was modeled as a discrete-time dynamical system, and the proposed control law was proven by Lyapunov arguments to drive the system state into an equilibrium condition that represents an approximation of the Wardrop user equilibrium. The limited availability of network resources was explicitly included in the control design, and the optimality of their usage was obtained in adversarial terms, as the various information flows compete with each other to maximise their connection performances.
All the researches included in this work covered different aspects and problems that arise when controlling a CPS. The common aspect that is shared among the four controllers is their focus on the safety and resiliency of the controlled systems, and the optimal usage of the limited available resources, in order to assure efficient and safe system operation and service provision.
In the first two researches this aspect is clear, since the optimal exploitation of the available resources (energy) was obtained by a centralised controller capable of predictive optimisation, and it is worth noting that the adversarial load balancing considered in the fourth case study still leads the network in a state in which all of its users cannot utilise better the available resources without cooperating. Even if paying the so-called Price of Anarchy, the load balancing attained can still improve the connection resiliency by enabling Multi-Connectivity (i.e., the routing of a single information flow over multiple paths), a core feature of 5G.
Furthermore, the direct improvement of spacecraft resiliency brought by life-support systems as the one discussed in this work (in principle capable even of reactivating a non-operative satellite), is paired with the optimal usage of the limited actuating capabilities typical of spacecrafts, potentially lowering significantly space missions cost.
The control methods utilised in this work are taken from different fields of Control Theory, mostly due to the fact that the problems studied were defined starting from applied research projects characterised by very heterogeneous requirements and goals. The rationale behind the choice of the various methodologies utilised is justified for each controller, and the tailoring of their characteristics to the specific applications is discussed in depth in their corresponding chapters
Distributed MARL with Limited Sensing for Robot Navigation Problems
Full text linkThis paper proposes a Multi-Agent Reinforcement Learning (MARL) algorithm for the multi-robot navigation problem. Most of the proposals in the literature requires some form of information sharing and communications among agents to coordinate their action in order to complete the overall task. The proposed paper, named Limited Sensing MARL (LS-MARL), assumes that each robot decisions rely on local information and is provided with sensor, which can be switched on for the localization of the robots within a given range. Besides the navigation task, each agent aims at limiting the use of the sensor as much as possible (i.e., to be as independent as possible) for energy saving or safety reasons. The algorithm is evaluated by simulations and favourably compares to the one proposed in (Yu et al. (2015)), that assumes a similar setup in which the neighbouring agents share their positioning information
AUTOMATION INTELLIGENCE AND CONTROL SRL
No full textSviluppo, produzione e commercializzazione di prodotti o servizi innovativi ad alto valore tecnologico.
Sviluppo, produzione e commercializzare di soluzioni basate su tecnologie di intelligenza artificiale, automazione e ottimizzazione, per l’analisi, la gestione, la valutazione e il controllo di sistemi e processi complessi al fine di migliorarne l’efficienza e la sostenibilità, anche integrando analisi dei dati e metodologie matematiche
Hierarchical Federated Learning for Edge Intelligence through Average Consensus
Full text linkConsensus of multi-agent systems has recently been studied in the context of Federated Learning (FL), an emerging branch of distributed machine learning. The present paper proposes a two-level hierarchical algorithm for FL in the context of edge computing, developing a fully decentralized solution that relies on results obtained for discrete-time consensus of dynamical systems. The proposed architecture and algorithm are validated on a test case and compared to current solutions, which require a centralized server
AdaFed: Performance-based Adaptive Federated Learning
Full text linkFederated Learning is a distributed and privacy-preserving machine learning technique that allows local clients to learn a model without sharing their own data by coordinating with a global server. In this work, we present the Adaptive Federated Learning (AdaFed) algorithm, which aims at improving the training performance of deep neural networks in Federated Learning settings by: (i) dynamically weighting the local models in the model averaging procedure; (ii) by adapting the loss function used by the federation at every communication round. We discuss the specialisation of AdaFed for both classification and regression tasks, providing several validation examples. Due to its adaptive design, the AdaFed algorithm showed a robust behaviour against unbalanced data distributions and adversarial clients
Deep reinforcement learning control of white-light continuum generation
Full text linkWhite-light continuum (WLC) generation in bulk media finds numerous applications in ultrafast optics and spectroscopy. Due to the complexity of the underlying spatiotemporal dynamics, WLC optimization typically follows empirical procedures. Deep reinforcement learning (RL) is a branch of machine learning dealing with the control of automated systems using deep neural networks. In this Letter, we demonstrate the capability of a deep RL agent to generate a long-term-stable WLC from a bulk medium without any previous knowledge of the system dynamics or functioning. This work demonstrates that RL can be exploited effectively to control complex nonlinear optical experiments
Automatic Transportation Mode Recognition on Smartphone Data Based on Deep Neural Networks
Full text linkIn the last few years, with the exponential diffusion of smartphones, services for turn-by-turn navigation have seen a surge in popularity. Current solutions available in the market allow the user to select via an interface the desired transportation mode, for which an optimal route is then computed. Automatically recognizing the transportation system that the user is travelling by allows to dynamically control, and consequently update, the route proposed to the user. Such a dynamic approach is an enabling technology for multi-modal transportation planners, in which the optimal path and its associated transportation solutions are updated in real-time based on data coming from (i) distributed sensors (e.g., smart traffic lights, road congestion sensors, etc.); (ii) service providers (e.g., car-sharing availability, bus waiting time, etc.); and (iii) the user’s own device, in compliance with the development of smart cities envisaged by the 5G architecture. In this paper, we present a series of Machine Learning approaches for real-time Transportation Mode Recognition and we report their performance difference in our field tests. Several Machine Learning-based classifiers, including Deep Neural Networks, built on both statistical feature extraction and raw data analysis are presented and compared in this paper; the result analysis also highlights which features are proven to be the most informative ones for the classification
Ensuring the Stability of Power Systems Against Dynamic Load Altering Attacks: A Robust Control Scheme Using Energy Storage Systems
Full text linkThis paper presents a robust protection scheme to protect the power transmission network against a class of feedback-based attacks referred in the literature as "Dynamic Load Altering Attacks" (D-LAAs). The proposed scheme envisages the usage of Energy Storage Systems (ESSs) to avoid the destabilising effects that a malicious state feedback has on the power network generators. The methodologies utilised are based on results from polytopic uncertain systems, invariance theory and Lyapunov arguments. Numerical simulations on a test scenario validate the proposed approach
User-aware centralized resource allocation in heterogeneous networks
Full text linkIn the last two years, in Europe, 5G networks and services proliferated. The integration of 5G networks with other radio access networks is considered one of the key enablers for matching the challenging 5G Quality of Service requirements. In particular, the integration with high throughput satellites promises to increase the network performances in terms of resilience and Quality of Service. The present work addresses this problem and presents a user-aware resource allocation methodology for heterogeneous networks. Said methodology is articulated in two-steps: at first, the Analytical Hierarchy Process is used for deciding the network over which traffic is steered and, then, a Cooperative Game for allocating resources within the network is set up. Simulations are presented for validating the proposed approach
- …
