Search

Orientadores: Paulo Sérgio Martins Pedro, Edson Luiz Ursini Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Tecnologia Resumo: Este trabalho procura solucionar um dos problemas mais críticos das redes it Flying AdHoc (FANET), que é a questão da coordenação da recarga de baterias de drones que voam em forma de enxame (Swarms). Para que as recargas sejam realizadas da melhor maneira possível, é desejável que o número de dispositivos de carregamento (estações base) não seja excessivamente alto principalmente devido ao alto custo de implantação. Por outro lado, também é necessário que, quando os drones quiserem recarregar, sempre haja uma fonte de energia disponível. Ou seja, buscamos o equilíbrio entre o número economicamente viável de estações de recarga e o abastecimento adequado dos drones quando isso for necessário. Para essa finalidade, propomos agentes (drones) munidos de inteligência interna, ou seja, com preditores internos que fornecem inteligência na tentativa de prever a próxima taxa de presença nesse dispositivo de carregamento e dessa maneira poder decidir se deve ou não ir abastecer. O ideal é que a previsão seja sempre a melhor possível. Assim, o drone deveria ir quando prevê que deve ir e não deveria ir quando prevê não ir. A utilização do equilíbrio de Nash para este problema, sendo possível pela utilização da modelagem via o Problema El Farol Bar (PEFB), permite o desenvolvimento de uma analogia sem o conluio dos agentes na coordenação da simulação da recarga desse conjunto de drones. Ou seja, com relação à coordenação do enxame para a recarga das baterias não haverá gasto de energia em comunicação, embora esta continue nas demais tarefas inerentes ao {\it swarm} de drones. A verificação da adequação da proposta é feita por meio de Simulação Baseada em Agentes e neste modelo são utilizadas três políticas diferentes para a tomada de decisão do melhor preditor para cada drone. Isso permite verificar o desempenho na operação do sistema por meio do Equilíbrio de Nash. No atual estado dessa analogia considera-se que, caso os drones forem para a estação de recarga e a mesma estiver cheia, não haverá recarga possível porque o sistema estará sobrecarregado. Neste estudo estão incluídas análises microscópicas e macroscópicas do desempenho do swarm. A análise microscópica é a avaliação do desempenho das recompensas de cada preditor em relação a um conjunto de parâmetros de simulação, visando um aumento da eficácia no desempenho microscópico. A análise macroscópica é a avaliação do desempenho do atendimento global do sistema com os três tipos de políticas. Esta última análise é usada como base para o desenvolvimento da analogia de recarga dos drones. Dessa maneira, avalia-se o desempenho dos melhores conjuntos de simulação para a recarga dos drones que permitem abastecer abaixo do limiar de recarga (número de posições disponível de recarga), mas que estão relativamente próximos a esse limiar Abstract: This work seeks to solve one of the most critical problems of the Flying AdHoc (FANET) networks, which is the issue of coordinating the recharging of drones that fly in the form of Swarms. For recharging to be done in the best possible way, it is desirable that the number of charging devices (base stations) did not be excessively high due to the high implementation cost. Conversely, it is also necessary that, when drones want to recharge, they must have a source of energy available. In other words, we search for a balance between the economically viable number of charging stations and the adequate energy supply for the drones when necessary. For this, we propose agents (drones) equipped with internal intelligence, that is, with internal predictors that provide intelligence to attempt to predict the next attendance rate in the charging device and thus be able to decide whether go or not go to the recharging. Ideally, the forecast should be as best as possible. Therefore, the drone should go when it predicts it should go and it shouldn't go when it predicts not to go. The Nash equilibrium usage for this problem is made possible by the modeling via the El Farol Bar Problem (EFBP), which allows the development of this analogy without the collusion of agents in coordinating the simulation of the recharge of this set of drones. In other words, there will be no energy expenditure on communication about the drones' battery recharging coordination, although the communication will continue in the other tasks inherent to the swarm of drones. The verification of the suitability of the proposal is done through Agent-Based Simulation and are used three different policies for the best predictor decision by each drone. This will allow us to verify the performance in the operation of the system through a Nash Equilibrium. In the current state of this analogy is considered that if the drones go to the charging station and it is full, there will be no possible charging because the system is overloaded. This study includes microscopic and macroscopic analysis. Microscopic analysis is the evaluation of the performance of the rewards of each predictor concerning a set of simulation parameters, aiming at a microscopic behavior performance improvement. A macroscopic analysis is the evaluation of the performance of the global service of the system with three types of policies. This latter analysis is used as a basis for evaluating the drone's recharge analogy. In this way, the performance of the best simulation sets for the recharge of drones is evaluated, which allows supplying below the control threshold (attendance below than the number of recharge positions), but which are relatively close to the threshold Mestrado Sistemas de Informação e Comunicação Mestre em Tecnologia CAPES 001

Although smart devices markets are increasing their sales figures, their computing capabilities are not sufficient to provide good-enough-quality services. This paper proposes a solution to organize the devices within the Cloud-Edge Continuum in such a way that each one, as an autonomous individual –Agent–, processes events/data on its embedded compute resources while offering its computing capacity to the rest of the infrastructure in a Function-as-a-Service manner. Unlike other FaaS solutions, the described approach proposes to transparently convert the logic of such functions into task-based workflows backing on task-based programming models; thus, agents hosting the execution of the method generate the corresponding workflow and offloading part of the workload onto other agents to improve the overall service performance. On our prototype, the function-to-workflow transformation is performed by COMPSs; thus, developers can efficiently code applications of any of the three envisaged computing scenarios – sense-process-actuate, streaming and batch processing – throughout the whole Cloud-Edge Continuum without struggling with different frameworks specifically designed for each of them. This work has been supported by the Spanish Government (PID2019-107255GB), by Generalitat de Catalunya (contract 2014-SGR-1051), and by the European Commission through the Horizon 2020 Research and Innovation program under Grant Agreement No. 101016577 (AI-SPRINT project).

Mobile applications have become a commodity in multiple daily scenarios. Their increasing complexity has led mobile software ecosystems to become heterogeneous in terms of hardware specifications, features and context of use, among others. For their users, fully exploiting their potential has become challenging. While enacting software systems with adaptation mechanisms has proven to ease this burden from users, mobile devices present specific challenges related to privacy and security concerns. Nevertheless, rather than being a limitation, users can play a proactive role in the adaptation loop by providing valuable feedback for runtime adaptation. To this end, we propose the use of chatbots to interact with users through a human-like smart conversational process. We depict a work-in-progress proposal of an end-to-end framework to integrate semi-automatic adaptation mechanisms for mobile applications. These mechanisms include the integration of both implicit and explicit user feedback for autonomous user categorization and execution of enactment action plans. We illustrate the applicability of such techniques through a set of scenarios from the Mozilla mobile applications suite. We envisage that our proposal will improve user experience by bridging the gap between users’ needs and the capabilities of their mobile devices through an intuitive and minimally invasive conversational mechanism. This work has been partially supported by AGAUR, code 2017-SGR-1694. The corresponding author gratefully acknowledges the Universitat Politècnica de Catalunya and Banco Santander for the financial support of his predoctoral grant FPI-UPC.

The advancement on explainability techniques is quite relevant in the field of Reinforcement Learning (RL) and its applications can be beneficial for the development of intelligent agents that are understandable by humans and are able cooperate with them. When dealing with Deep RL some approaches already exist in the literature, but a common problem is that it can be tricky to define whether the explanations generated for an agent really reflect the behaviour of the trained agent. In this work we will apply an approach for explainability based on the creation of a Policy Graph (PG) that represents the agent’s behaviour. Our main contribution is a way to measure the similarity between the explanations and the agent’s behaviour, by building another agent that follows a policy based on the explainability method and comparing the behaviour of both agents.

Path finding for autonomous agents has been traditionally driven by finding optimal paths, typically by using A* search or any of its variants. When it comes to simulating virtual humanoids, traditional approaches rarely consider aspects of human memory or orientation. In this work, we propose a new path finding algorithm, inspired by current research regarding how the brain learns and builds cognitive maps. Our method represents the space as a hexagonal grid with counters, based on brain research that has investigated how memory cells are fired. Our path finder then combines a method for exploring unknown environments while building such a cognitive map, with an A* search using a modified heuristic that takes into account the cognitive map. The resulting paths show how as the agent learns the environment, the paths become shorter and more consistent with the optimal A* search. Moreover, we run a perceptual study to demonstrate that the viewers could successfully identify the intended level of knowledge of the simulated agents. This line of research could enhance the believability of autonomous agents’ path finding in video games and other VR applications. This work was partly funded by the Spanish Ministry of Economy, Industry and Competitiveness under Grant No. TIN2017- 88515-C2-1-R.

There may be a lot of abusive behaviour in conversations between teenagers, which take place through social media. In this project, we develop classifiers to find out which texts present abuse such as violence, sexual behaviour, disorder, anxiety... For this reason, we use different classifiers that take advantage of the most recent techniques based on language modeling using deep neural networks, namely Transformers. But we find it difficult to obtain supervised data, i.e. data tagged with the feeling it contains. Luckily, thanks to the countless sources of information available on the Internet, we can get unsupervised data and examples of text on the Internet. Consequently, this project aims to explore the possibility of using unsupervised data, beyond using a supervised set to train a system in the task of sentiment classification. More specifically in short texts, in informal contexts and in Catalan and Spanish. We have a set of about 200,000 supervised samples and a set of 300,000 unsupervised samples. We will introduce and study different models of Transformers (BERT, GAN- BERT, XLM-RoBERTa and ELECTRA) and their performance in the work of sentiment analysis. To finish with, based on models that achieve a better score in the binary classification task of feelings, i.e. with ELECTRA and XLM-RoBERTa, we construct a multi- label classifier. En las conversaciones entre adolescentes, que se dan por medio de las redes sociales, pueden existir muchos comportamientos abusivos. En este proyecto desarrollamos clasificadores que detecten qué textos presentan algún abuso como violencia, conductas sexuales, desórdenes, ansiedad... Por este motivo, usamos diferentes clasificadores que aprovechan las técnicas más recientes basadas en modelado de lenguaje mediante redes neuronales profundas, concretamente Transformers. Pero nos encontramos con la dificultad de obtener datos supervisados, es decir datos etiquetados con el sentimiento que contiene. Por suerte, gracias a la infinidad de fuentes de información de las que disponemos en la red, podemos conseguir datos y ejemplos de texto no supervisado a internet. Por consiguiente, este proyecto quiere explorar la posibilidad de utilizar datos no supervisados, más allá de usar un conjunto de datos supervisado para entrenar un sistema en la tarea de clasificación de sentimientos. Más concretamente en textos cortos, en contextos informales y de lengua catalana y castellana. Contamos con un set de cerca de 200.000 muestras supervisadas y un set de 300.000 muestras no supervisadas. Introduciremos y estudiaremos diferentes modelos de Transformers (BERT, GAN-BERT, XLM-RoBERTa y ELECTRA) y su rendimiento en la tarea de análisis de sentimientos. Finalmente, a partir de los modelos que consiguen una mejor puntuación en la tarea de clasificación binaria de sentimientos, es decir con ELECTRA y XLM-RoBERTa, construimos un clasificador multi-etiqueta. En les converses entre adolescents, que es donen per mitjà de les xarxes socials, poden existir molts comportaments abusius. En aquesta tesi desenvolupem classificadors que detectin quins textos presenten algun abús com violència, conductes sexuals, desordres, ansietat... Per aquest motiu, faig servir diferents classificadors que aprofiten les tècniques més recents basades en modelat de llenguatge mitjançant xarxes neuronals profundes, concretament Transformers. Però ens trobem amb la dificultat d?obtenir dades supervisades, és a dir dades etiquetades amb el sentiment que conté. Per sort, gràcies a la infinitat de fonts d?informació de les que disposem a la xarxa, podem aconseguir dades i exemples de text no supervisat a internet. Per consegüent, aquest projecte vol explorar la possibilitat d'utilitzar dades no supervisades, més enllà de fer servir un set supervisat per entrenar un sistema en la tasca de classificació de sentiments. Més concretament en textos curts, en contextos informals i de llengua catalana i castellana. Comptem amb un set de prop de 200.000 mostres supervisades i un set de 300.000 mostres no supervisades. Introduirem i estudiarem diferents models de Transformers (BERT,GAN-BERT,XLM-Roberta i Electra) i el seu rendiment en la tasca d'anàlisi de sentiments. Finalment, a partir dels models que aconsegueixen una millor puntuació en la tasca de classificació binària de sentiments, és a dir amb Electra i XLM-Roberta, construïm un classificador multi-label

Hospital patient waiting times and length of stay are indicators of the quality of emergency department (ED) services, factors that are a ected by the number of patient arrivals. It is necessary to ac- curately estimate ED patient arrivals in order to manage resources e ectively. Prediction models, however, are conditioned by the hospi- tal population and its placement (i.e. meteorological conditions). In the particular case of a tourist region, the population has an impor- tant amount of variability which challenge EDs. This paper aims to address ED attendances predictions for an hospital placed in a tourist region by means of a new approach that combines multiple linear re- gression with arti cial neural networks and regression tree ensembles, looking for dealing for ED variability and prediction for a week time horizon that enables operational reaction to the ED responsible. The methodology uses exogenous variables such as calendar, weather and socio-economic data to improve the accuracy of these forecasts. Pre- diction models are built on data for 11-years and the predictions are tested over 1-year. The results showed that the proposed methodol- ogy is capable to perform weekly predictions with an error about 5%, demonstrating that it could be used by EDs The work was carried out with support from the Generalitat de Catalunya 2017 SGR 1551

O processo de refatoração garante uma qualidade maior no código-fonte aumentando a sua manutenibilidade, confiabilidade e flexibilidade. De acordo com a literatura cerca de 70% do custo do software é para manutenção e isto pode ser diminuído usando técnicas que permitem aumentar a qualidade do código-fonte tal como a refatoração baseada em padrões de projetos. Constatou-se por meio de um mapeamento sistemático que os trabalhos na literatura de detecção e inserção de padrões de projeto não são realizados de forma autônoma. Este trabalho criou a abordagem Codice-Unio para detectar pontos de inserção e aplicar padrões de projeto com agentes usando a arquitetura de Belief-Desire-Intention (BDI). A abordagem contempla em um mesmo ambiente três métodos da literatura capazes de detectar e aplicar padrões de projeto em código-fonte escrito em linguagem Java. A fim de comparar o processo de refatoração antes e depois da aplicação do padrão de projeto foi contemplado na abordagem a avaliação de métricas relacionadas aos atributos de qualidade tais como manutenibilidade, reusabilidade e confiabilidade. A Codice-Unio foi implementada em um framework para agentes que suporta a arquitetura BDI e usou ferramentas específicas para leitura de código e avaliação dos atributos de qualidade. A abordagem foi avaliada com cenários de testes providos pelos métodos da literatura e posteriormente por projetos open-source encontrados na web via GitHub. Como resultado, a Codice-Unio é capaz de identificar e aplicar padrões de projeto em classes candidatas automaticamente em cerca de aproximadamente 97% dos projetos que foram submetidos ao experimento. The refactoring process guarantees a higher quality in the source code, increasing its maintainability, readability and flexibility. According to the literature about 70% of the cost of the software is for maintenance and this can be reduced using techniques that allow to increase the quality of the source code such as refactoring based on design standards. It was found through a systematic mapping that the works in the literature for the detection and insertion of design patterns are not carried out autonomously. This work created the Codice-Unio approach to detect insertion points and apply design patterns with agents using the Belief-Desire-Intention (BDI) architecture. The approach contemplates in the same environment three methods from the literature capable of detecting and applying design patterns in source code written in Java language. In order to compare the refactoring process before and after the application of the design standard, the assessment of metrics related to quality attributes such as maintainability, reusability and readability was contemplated in the approach. CodiceUnio was implemented in an agent framework that supports the BDI architecture and used specific tools for reading code and evaluating quality attributes. The approach was evaluated using test scenarios provided by the literature methods and later by open-source projects found on the web via GitHub. As a result, Codice-Unio is able to automatically identify and apply design patterns in candidate classes in approximately 97% of the projects that have undergone the experiment.

La empresa Bookline ha desarrollado un asistente virtual que puede atender llamadas a restaurantes y ayudar a los usuarios a realizar una reserva. Este trabajo pretende implementar todas sus funcionalidades a través de un chatbot que permita al asistente atender a usuarios a través de los servicios de mensajería instantánea más populares. El objetivo és diseñar un programa que pueda atender a los usuarios a través de chatbots de Whatsapp, Telegram y Facebook Messenger, y que se pueda extender fácilmente a otros servicios insertando unas mínimas modificaciones. Para el desarrollo de este proyecto se han utilizado principalmente servicios de Google Cloud Platform y el lenguaje de programación Python. El resultado final del proyecto és un chatbot para cada servicio de mensajería que permita dialogar en castellano o inglés, capaz de responder a cualquier pregunta igual que lo haría el asistente de Bookline y que permita reservar una mesa en los restaurantes que contraten este servicio. The company Bookline has developed a virtual assistant that attends calls to restaurants and helps users at making reservations. This project aims to implement all of the assistant's functions in a chatbot that would allow the assistant to attend users through the most popular instant messaging services. My objective is to design a program capable of attending users through chatbots in Whatsapp, Telegram and Facebook Messenger, and make it easy to extend it to other services through some minimal modifications. For the development of this project the tools that have been primarily used were some services from Google Cloud Platform and the programming language Python. The project's final result is a chatbot for each messaging service that can follow a conversation in Spanish or English, capable of responding to any question as Bookline's assistant would, and capable of booking a table in any restaurant that has hired this service. L'empresa Bookline ha desenvolupat un assistent virtual que pot atendre les trucades a restaurants i ajudar als usuaris a realitzar una reserva. Aquest treball pretén implementar totes les seves funcions a través d'un chatbot que permeti a l'assistent atendre als usuaris a través dels serveis de missatgeria instantània més populars. L'objectiu és dissenyar un programa que pugui atendre als usuaris a través de chatbots de Whatsapp, Telegram i Facebook Messenger, i que es pugui estendre fàcilment a altres serveis fent les modificacions mínimes. Per al desenvolupament d'aquest projecte s'han fet servir principalment serveis de Google Cloud Platform i el llenguatge de programació Python. El resultat final del projecte és un chatbot per a cada servei de missatgeria que permet dialogar amb castellà o anglès, capaç de respondre a qualsevol pregunta igual que l'assistent de Bookline i que permeti reservar una taula en els restaurants que contractin aquest servei.

Els models BPMN (Business Process Modeling and Notation), els quals permeten representar els processos de negoci d'una empresa, no resulten fàcils d'entendre per part d'aquelles persones no qualificades. En l'àmbit laboral, la majoria de treballadors no tenen els coneixements per entendre'ls. Això fa que es requereixi algú que expliqui el funcionament del negoci. A més, no sempre és possible tenir algú a qui preguntar un dubte, el que pot causa que un treballador no sàpiga que fer. La problemàtica dels BPMN també pot aplicar-se a treballadors que si entenen els models. Per exemple, un alt càrrec, que vulgui consultar un detall d'un determinat procés, pot perdre molt de temps buscant entre molts models per tal de trobar el que busca. L'objectiu general del projecte és crear una aplicació que transformi processos de negoci en assistents virtuals. Concretament, el projecte construeix un prototip fent ús de les tecnologies més punteres i avalua les tecnologies segons la flexibilitat del llenguatge obtinguda. En aquest context, la flexibilitat del llenguatge s'entén com a la diversitat del vocabulari, que pot utilitzar una persona, en referir-se a un mateix concepte. El prototip selecciona un model BPMN, el processa i genera frases de resposta. A més, genera frases amb significat similar, per tal d'augmentar la flexibilitat del llenguatge. Un cop acabat el processat, s'utilitza Dialogflow (una eina de modelatge d'assistents virtuals) per tal de construir l'assistent virtual. Finalment i durant una conversa amb l'assistent, s'utilitza un servidor per a emmagatzemar la informació pertinent. Tenint en compte el prototip construït, el projecte conclou que dialogflow és una bona eina pel que fa a la generació d'assistents virtuals a partir d'un model BPMN. Tot i això, si es vol aconseguir un bon resultat en la flexibilitat del llenguatge, s'ha de dedicar grans esforços en la generació de frases amb significat similar. El projecte conclou que es requereix més recerca per aconseguir una major flexibilitat en el llenguatge Business Process Modeling and Notation (BPMN) models, which allow you to represent the business processes of a company, are not easy to understand by those who are not qualified. In the workplace, most workers do not have the knowledge to understand them. This requires someone to explain how the business works. Also, it is not always possible to have someone to ask a question to, which can cause a worker to not know what to do. The problem of BPMNs can also be applied to workers who do understand the models. For example, a senior official, who wants to consult a detail of a certain process, may waste a lot of time searching among many models in order to find what he is looking for. The overall goal of the project is to create an application that transforms business processes into virtual assistants. Specifically, the project builds a prototype using the most advanced technologies and evaluates the technologies according to the flexibility of the language obtained. In this context, the flexibility of language is understood as the diversity of vocabulary, which a person can use, when referring to the same concept. The prototype selects a BPMN model, processes it, and generates response sentences. In addition, it generates sentences with similar meanings, in order to increase the flexibility of the language. Once the processing is complete, Dialogflow (a virtual assistant modeling tool) is used to build the virtual assistant. Finally, during a conversation with the wizard, a server is used to store the relevant information. Given the prototype built, the project concludes that dialogflow is a good tool for generating virtual assistants from a BPMN model. However, if you want to get a good result in language flexibility, you have to put a lot of effort into generating sentences with similar meaning. The project concludes that more research is required to achieve greater flexibility in language. Los modelos BPMN (Business Process Modeling and Notation), los cuales permiten representar los procesos de negocio de una empresa, no resultan fáciles de entender por parte de aquellas personas no cualificadas. En el ámbito laboral, la mayoría de trabajadores no tienen los conocimientos para entenderlos. Esto hace que se requiera alguien que explique el funcionamiento del negocio. Además, no siempre es posible tener a alguien a quien preguntar una duda, lo que puede causa que un trabajador no sepa que hacer. La problemática de los BPMN también puede aplicarse a trabajadores que si entienden los modelos. Por ejemplo, un alto cargo, que quiera consultar un detalle de un determinado proceso, puede perder mucho tiempo buscando entre muchos modelos a fin de encontrar lo que busca. El objetivo general del proyecto es crear una aplicación que transforme procesos de negocio en asistentes virtuales. Concretamente, el proyecto construye un prototipo en uso de las tecnologías más punteras y evalúa las tecnologías según la flexibilidad del lenguaje obtenida. En este contexto, la flexibilidad del lenguaje se entiende como la diversidad del vocabulario, que puede utilizar una persona, al referirse a un mismo concepto. El prototipo selecciona un modelo BPMN, lo procesa y genera frases de respuesta. Además, genera frases con significado similar, a fin de aumentar la flexibilidad del lenguaje. Una vez terminado el procesado, se utiliza Dialogflow (una herramienta de modelado de asistentes virtuales) para construir el asistente virtual. Finalmente y durante una conversación con el asistente, se utiliza un servidor para almacenar la información pertinente. Teniendo en cuenta el prototipo construido, el proyecto concluye que dialogflow es una buena herramienta en cuanto a la generación de asistentes virtuales a partir de un modelo BPMN. Sin embargo, si se quiere conseguir un buen resultado en la flexibilidad del lenguaje, se debe dedicar grandes esfuerzos en la generación de frases con significado similar. El proyecto concluye que se requiere de más investigación para conseguir una mayor flexibilidad en el lenguaje.

Our work is concerned with user modelling in open environments. Our proposal then is the line of contributions to the advances on user modelling in open environments thanks so the Agent Technology, in what has been called Smart User Model. Our research contains a holistic study of User Modelling in several research areas related to users. We have developed a conceptualization of User Modelling by means of examples from a broad range of research areas with the aim of improving our understanding of user modelling and its role in the next generation of open and distributed service environments. This report is organized as follow: In chapter 1 we introduce our motivation and objectives. Then in chapters 2, 3, 4 and 5 we provide the state-of-the-art on user modelling. In chapter 2, we give the main definitions of elements described in the report. In chapter 3, we present an historical perspective on user models. In chapter 4 we provide a review of user models from the perspective of different research areas, with special emphasis on the give-and-take relationship between Agent Technology and user modelling. In chapter 5, we describe the main challenges that, from our point of view, need to be tackled by researchers wanting to contribute to advances in user modelling. From the study of the state-of-the-art follows an exploratory work in chapter 6. We define a SUM and a methodology to deal with it. We also present some cases study in order to illustrate the methodology. Finally, we present the thesis proposal to continue the work, together with its corresponding work scheduling and temporalisation

Low-cost low-margin implementation plays an essential role in upgrading optical metro networks required for future 5G ecosystem. In this regard, low-resolution analog-to-digital converters can be used in coherent optical transponders to reduce cost and power consumption. However, the resulting transmission systems become more sensitive to physical layer fluctuations like the events caused by fiber stressing. Such fluctuations might have a strong impact on the quality of transmission (QoT) of the signals. To guarantee robust operation, soft decision forward error correction (FEC) techniques are required to guarantee zero post-FEC bit error rate (BER) transmission, which could increase the power consumption of the receiver and thus operational expenses. In this paper, we aim at minimizing power consumption while keeping zero post-FEC errors by means of a predictive autonomic transmission agent (ATA) based on machine learning. We present a sophisticated ATA model that, taking advantage of real-time monitoring of state of polarization traces and the corresponding pre-FEC BER, predicts the right FEC configuration for short-term operation, thus requiring minimum power consumption. In addition, we propose a complementary long-term prediction of excessive pre-FEC BER to enable remote reconfiguration at the transmitter side through the network controller. A set of experimental measurements is used to train and validate the proposed ATA system. Exhaustive numerical analysis allows concluding that ATA based on artificial neural network predictors achieves the maximum QoT robustness with 80% power consumption reductions compared to static FEC configuration. The research leading to these results has received funding from the European Commission for the H2020-ICT-2016-2 METRO-HAUL project (G.A. 761727), from the AEI/FEDER TWINS project (TEC2017-90097-R), and from the Catalan Institution for Research and Advanced Studies (ICREA).

[EN] Existing HW/SW platforms for safety-critical systems suffer from limited performance and/or from lack of flexibility due to building on specific proprietary components. This jeopardizes their wide deployment across domains. While some research has been done to overcome these limitations, they have had limited success owing to missing flexibility and extensibility. Flexibility and extensibility are the cornerstones of industry adoption: industries dealing in capital goods need technologies on which they can rely on during decades (e.g. avionics, space, automotive). SELENE aims at covering this gap by proposing a new family of safety-critical computing platforms, which builds upon open source components such as the RISC-V instruction set architecture, GNU/Linux, and the Jailhouse hypervisor. SELENE will develop an advanced computing platform that is able to: (1) adapt the system to the specific requirements of different application domains, to changing environmental conditions, and to internal conditions of the system itself; (2) allow the integration of applications of different criticalities and performance demands in the same platform, guaranteeing functional and temporal isolation properties; (3) achieve flexible diverse redundancy by exploiting the inherent redundant capabilities of the multicore; and (4) efficiently execute compute-intensive applications by means of specific accelerators., This work has received funding from the European Unions Horizon 2020 research and innovation programme under grant agreement no. 871467.