Continual learning on nano drones

Miniaturized flying robotic platforms, called nano-drones, have the potential to revolutionize the autonomous robots industry sector thanks to their very small form factor. The nano-drones’ limited payload only allows for a sub-100mW microcontroller unit for the on-board computations. Therefore, traditional computer vision and control algorithms are too computationally expensive to be executed on board these palm-sized robots, and we are forced to rely on artificial intelligence to trade off accuracy in favor of lightweight pipelines for autonomous tasks. However, relying on deep learning exposes us to the problem of generalization since the deployment scenario of a convolutional neural network (CNN) is often composed by different visual cues and different features from those learned during training, leading to poor inference performances. Our objective is to develop and deploy and adaptation algorithm, based on the concept of latent replays, that would allow us to fine-tune a CNN to work in new and diverse deployment scenarios. To do so we start from an existing model for visual human pose estimation, called PULPFrontnet, which is used to identify the pose of a human subject in space through its 4 output variables, and we present the design of our novel adaptation algorithm, which features automatic data gathering and labeling and on-device deployment. We therefore showcase the ability of our algorithm to adapt PULP-Frontnet to new deployment scenarios, improving the R2 scores of the four network outputs, with respect to an unknown environment, from approximately [−0.2, 0.4, 0.0,−0.7] to [0.25, 0.45, 0.2, 0.1]. Finally we demonstrate how it is possible to fine-tune our neural network in real time (i.e., under 76 seconds), using the target parallel ultra-low power GAP 8 System-on-Chip on board the nano-drone, and we show how all adaptation operations can take place using less than 2mWh of energy, a small fraction of the available battery power.

LOW COST ANALYZER FOR THE DETERMINATION OF PHOSPHORUS BASED ON OPEN-SOURCE HARDWARE AND PULSED FLOWS

The need for automated analyzers for industrial and environmental samples has triggered the research for new and cost-effective strategies of automation and control of analytical systems. The widespread availability of open-source hardware together with novel analytical methods based on pulsed flows have opened the possibility of implementing standalone automated analytical systems at low cost. Among the areas that can benefit from this approach are the analysis of industrial products and effluents and environmental analysis. In this work, a multi-pumping flow system is proposed for the determination of phosphorus in effluents and polluted water samples. The system employs photometric detection based on the formation of molybdovanadophosphoric acid, and the fluidic circuit is built using three solenoid micropumps. The detection is implemented with a low cost LED-photodiode photometric detection system and the whole system is controlled by an open-source Arduino Uno microcontroller board. The optimization of the timing to ensure the color development and the pumping cycle is discussed for the proposed implementation. Experimental results to evaluate the system behavior are presented verifying a linear relationship between the relative absorbance and the phosphorus concentrations for levels as high as 50 mg L-1.

Implementazione di una rete di sensori per il controllo di gas inquinanti nell'ambiente

Negli ultimi anni i progressi tecnologici in termini di miniaturizzazione elettronica, hanno permesso la realizzazione di componenti hardware ed in particolare di microprocessori e sensori dalle dimensioni ridottissime. Questo ha favorito la recente diffusione di reti di sensori wireless (Wireless Sensor Network) basate su sistemi embedded più o meno complessi ed applicate a settori di mercato che vanno dalla domotica alle applicazioni industriali, fino al monitoraggio dei pazienti. Lo scopo di questa tesi, svolta in collaborazione con la società Rinnova di Forlì, consiste nell’implementazione di un dimostratore che mostri la reale capacità di realizzare una rete WS che si appoggia su di un sistema embedded commerciale ed ampiamente diffuso come la piattaforma Arduino ed in grado di rilevare il livello di ammoniaca presente negli allevamenti di pollame. Tale gas infatti, se presente in quantità notevole, provoca una dannosa alterazione comportamentale dei polli e risulta quindi un parametro molto importante da monitorare. Oltre al sensore di ammoniaca, misurazione principale richiesta dal progetto, ne sono stati aggiunti uno per la temperatura ed uno per l’umidità. L’architettura finale implementata è quella tipica di una rete a stella, in cui il master centrale colleziona a polling i dati provenienti dai sensori collegati agli slave e li invia ad un server web, rendendoli accessibili mediante la rete Internet. L’utente finale può così accedere alla pagina web da un qualunque PC dotato di connessione Internet, monitorare i dati dei sensori e soprattutto verificare quando il livello di ammoniaca supera la soglia di attenzione, potendo così intervenire immediatamente nell’allevamento per effettuare le dovute operazioni di pulizia.

Reutilização de baterias automotivas como fonte alternativa de energia.

Atualmente um dos principais objetivos na área de pesquisa tecnológica é o desenvolvimento de soluções em favor do Meio Ambiente. Este trabalho tem por objetivo demonstrar a reutilização e consequentemente o aumento da vida útil de uma bateria Chumbo-Ácido, comumente instaladas em veículos automóveis, bem como beneficiar locais e usuários remotos onde o investimento na instalação de linhas de transmissão se torna inviável geográfica e economicamente, utilizando a luz solar como fonte de energia. No entanto a parte mais suscetível a falhas são as próprias baterias, justamente pela vida útil delas serem pequenas (em torno de 3 anos para a bateria automotiva) em comparação com o restante do sistema. Considerando uma unidade que já foi usada anteriormente, a possibilidade de falhas é ainda maior. A fim de diagnosticar e evitar que uma simples bateria possa prejudicar o funcionamento do sistema como um todo, o projeto considera a geração de energia elétrica por células fotovoltaicas e também contempla um sistema microcontrolado para leitura de dados utilizando o microcontrolador ATmega/Arduino, leitura de corrente por sensores de efeito hall da Allegro Systems, relés nas baterias para abertura e fechamento delas no circuito e um sistema de alerta para o usuário final de qual bateria está em falha e que precisa ser reparada e/ou trocada. Esse projeto foi montado na Ilha dos Arvoredos SP, distante da costa continental em aproximadamente 2,0km. Foram instaladas células solares e um banco de baterias, a fim de estudar o comportamento das baterias. O programa pôde diagnosticar e isolar uma das baterias que estava apresentando defeito, a fim de se evitar que a mesma viesse a prejudicar o sistema como um todo. Por conta da dificuldade de locomoção imposta pela geografia, foi escolhido o cartão SD para o armazenamento dos dados obtidos pelo Arduino. Posteriormente os dados foram compilados e analisados. A partir dos resultados apresentados podemos concluir que é possível usar baterias novas e baterias usadas em um mesmo sistema, de tal forma que se alguma das baterias apresentar uma falha o sistema por si só isolará a unidade.

Methodology for developing a Speech into Sign Language Translation System in a New Semantic Domain

This paper proposes a methodology for developing a speech into sign language translation system considering a user-centered strategy. This method-ology consists of four main steps: analysis of technical and user requirements, data collection, technology adaptation to the new domain, and finally, evalua-tion of the system. The two most demanding tasks are the sign generation and the translation rules generation. Many other aspects can be updated automatical-ly from a parallel corpus that includes sentences (in Spanish and LSE: Lengua de Signos Española) related to the application domain. In this paper, we explain how to apply this methodology in order to develop two translation systems in two specific domains: bus transport information and hotel reception.

Schizophrenic Psychoses with Motor-Dominant Symptoms: Considerations for a System-Specific Approach

In order to reduce heterogeneity in schizophrenia, a system-specific approach consisting of the domains "language", "affect" and "motor behavior" has been proposed. We examined this system-specific approach for its applicability to clinical practice in the motor behavior domain, using the methodological approach of case studies, and discuss here the differences to the positive/negative concept. We analyzed eight cases with stable motor-dominant symptoms, and also quantitatively assessed motor behavior by using the Bern Psychopathology Scale (BPS), a standardized psychopathological assessment instrument, as well as actigraphic data. Characterization of cases using the positive/negative approach was not helpful. We found an overlap of the motor behavior domain with the other two domains. This complicates the application of the system-specific approach in the sense of a typology. Furthermore, we found both relapsing courses with full remission and chronic courses with deterioration within the motor-dominant subtype. Nevertheless, the system-specific approach has heuristic utility for the future.

Factored Translation Models for improving a Speech into Sign Language Translation System

This paper proposes the use of Factored Translation Models (FTMs) for improving a Speech into Sign Language Translation System. These FTMs allow incorporating syntactic-semantic information during the translation process. This new information permits to reduce significantly the translation error rate. This paper also analyses different alternatives for dealing with the non-relevant words. The speech into sign language translation system has been developed and evaluated in a specific application domain: the renewal of Identity Documents and Driver’s License. The translation system uses a phrase-based translation system (Moses). The evaluation results reveal that the BLEU (BiLingual Evaluation Understudy) has improved from 69.1% to 73.9% and the mSER (multiple references Sign Error Rate) has been reduced from 30.6% to 24.8%.

Adapting a speech into sign language translation system to a new domain

This paper presents a methodology for adapting an advanced communication system for deaf people in a new domain. This methodology is a user-centered design approach consisting of four main steps: requirement analysis, parallel corpus generation, technology adaptation to the new domain, and finally, system evaluation. In this paper, the new considered domain has been the dialogues in a hotel reception. With this methodology, it was possible to develop the system in a few months, obtaining very good performance: good speech recognition and translation rates (around 90%) with small processing times.

A novel approach to antigen-specific deletion of CTL with minimal cellular activation using alpha 3 domain mutants of MHC class I/peptide complex

In this study, we have compared the effector functions and fate of a number of human CTL clones in vitro or ex vivo following contact with variant peptides presented either on the cell surface or in a soluble multimeric format. In the presence of CD8 coreceptor binding, there is a good correlation between TCR signaling, killing of the targets, and Fast-mediated CTL apoptosis. Blocking CD8 binding using (alpha3 domain mutants of MHC class I results in much reduced signaling and reduced killing of the targets. Surprisingly, however, Fast expression is induced to a similar degree on these CTLs, and apoptosis of CTL is unaffected. The ability to divorce these events may allow the deletion of antigen-specific and pathological CTL populations without the deleterious effects induced by full CTL activation.

A framework for the development of parallel and distributed real-time embedded systems

Embedded real-time applications increasingly present high computation requirements, which need to be completed within specific deadlines, but that present highly variable patterns, depending on the set of data available in a determined instant. The current trend to provide parallel processing in the embedded domain allows providing higher processing power; however, it does not address the variability in the processing pattern. Dimensioning each device for its worst-case scenario implies lower average utilization, and increased available, but unusable, processing in the overall system. A solution for this problem is to extend the parallel execution of the applications, allowing networked nodes to distribute the workload, on peak situations, to neighbour nodes. In this context, this report proposes a framework to develop parallel and distributed real-time embedded applications, transparently using OpenMP and Message Passing Interface (MPI), within a programming model based on OpenMP. The technical report also devises an integrated timing model, which enables the structured reasoning on the timing behaviour of these hybrid architectures.

Poster: EMMON: a WSN system architecture and toolset for large-scale and dense real-time embedded monitoring

Wireless sensor networks (WSNs) have attracted growing interest in the last decade as an infrastructure to support a diversity of ubiquitous computing and cyber-physical systems. However, most research work has focused on protocols or on specific applications. As a result, there remains a clear lack of effective and usable WSN system architectures that address both functional and non-functional requirements in an integrated fashion. This poster outlines the EMMON system architecture for large-scale, dense, real-time embedded monitoring. It provides a hierarchical communication architecture together with integrated middleware and command and control software. It has been designed to maintain as much as flexibility as possible while meeting specific applications requirements. EMMON has been validated through extensive analytical, simulation and experimental evaluations, including through a 300+ nodes test-bed the largest single-site WSN test-bed in Europe.

EMMON: a WSN system architecture for large scale and dense real-time embedded monitoring

Wireless sensor networks (WSNs) have attracted growing interest in the last decade as an infrastructure to support a diversity of ubiquitous computing and cyber-physical systems. However, most research work has focused on protocols or on specific applications. As a result, there remains a clear lack of effective, feasible and usable system architectures that address both functional and non-functional requirements in an integrated fashion. In this paper, we outline the EMMON system architecture for large-scale, dense, real-time embedded monitoring. EMMON provides a hierarchical communication architecture together with integrated middleware and command and control software. It has been designed to use standard commercially-available technologies, while maintaining as much flexibility as possible to meet specific applications requirements. The EMMON architecture has been validated through extensive simulation and experimental evaluation, including a 300+ node test-bed, which is, to the best of our knowledge, the largest single-site WSN test-bed in Europe to date.

Time-domain optimization of amplifiers based on distributed genetic algorithms

Thesis presented in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the subject of Electrical and Computer Engineering

Vision-based portuguese sign language recognition system

Vision-based hand gesture recognition is an area of active current research in computer vision and machine learning. Being a natural way of human interaction, it is an area where many researchers are working on, with the goal of making human computer interaction (HCI) easier and natural, without the need for any extra devices. So, the primary goal of gesture recognition research is to create systems, which can identify specific human gestures and use them, for example, to convey information. For that, vision-based hand gesture interfaces require fast and extremely robust hand detection, and gesture recognition in real time. Hand gestures are a powerful human communication modality with lots of potential applications and in this context we have sign language recognition, the communication method of deaf people. Sign lan- guages are not standard and universal and the grammars differ from country to coun- try. In this paper, a real-time system able to interpret the Portuguese Sign Language is presented and described. Experiments showed that the system was able to reliably recognize the vowels in real-time, with an accuracy of 99.4% with one dataset of fea- tures and an accuracy of 99.6% with a second dataset of features. Although the im- plemented solution was only trained to recognize the vowels, it is easily extended to recognize the rest of the alphabet, being a solid foundation for the development of any vision-based sign language recognition user interface system.

Static and dynamic interpretations of hand specific body language cues for HCI

Otto-von-Guericke-Universität Magdeburg, Fakultät für Elektrotechnik und Informationstechnik, Univ., Dissertation, 2015