Microbial fuel cell stacks (MFC-Stack) are often confronted with voltage reversals, likely due to an interplay between microbial community dynamics and insufficient electric circuit balancing. Herein, we provide new insight into voltage reversals by examining the microbiomes of twelve MFC units of a 12-liter Pilot-MFC-Stack during repair. Different biofilm repair methods (self-healing, electrostimulation, and re-acclimatization upon cross-inoculation) were used to evaluate the microbial community response. In addition, MFC-Stack simulation was performed based on Kirchhoff’s Second Law to predict values for source potentials and post-evaluate internal resistances. Analysis of the 16S rRNA amplicon sequencing data suggests that the biofilm repair methods could slowly heal damaged biofilms. Notably, severely voltage reversed MFC units had low electrogen relative abundances (18%) and positive anode potentials, while strong bioanodes and contained more than 50% electrogens and had negative anode potentials. Between-community analyses (beta diversity ordination and multinomial regression) of the voltage reversed MFC units revealed differences among biofilms in contrast to healthy/strong MFC units. Permutational multivariate analysis of variance (PERMANOVA) confirmed that reversed biofilms were, indeed, significantly (p < 0.05) different from stronger ones. Overall, these analyses demonstrated the utility of combining electrotechnical and microbial community analyses, especially beta diversity ordination and multinomial regression, to understand problematic MFC units and the potential success of a biofilm repair method. Finally, thicker biofilms were usually healthier and stronger, although thickness was no guarantee for proper structure and power function as all factors were interdependent. There was an evolutionary trend that strong anodes became stronger/healthier and others weaker. This spontaneous trend has to be considered to avoid irreversible voltage reversals and to repair electrogenic biofilms in an MFC-Stack.
This article examines the entanglement between feelings of stress and discomfort, physiological arousal and urban experiences of persons living with early psychosis. It adopts a biosocial approach, using mixed methods combining ambulatory skin conductance monitoring, mobile interviews and contextual data, collected through GPS and video recordings. The study draws on and strives to cross-fertilize two recent strands of research. The first relates to the use of digital phenotyping in mental health research. The second explores stress and emotional arousal in cities using ambulatory physiological measures. Empirically, the paper is based on fieldwork in Basel, Switzerland, with nine participants recruited within the Basel Early Treatment Service (BEATS), and four controls. We focus on three salient elements in our results: visual perception of moving bodies, spatial transitions and openness and enclosure of the built environment. The analysis shows how these elements elicit physiological responses of arousal and expressed feelings of discomfort. In the concluding section we discuss the methodological implications of these results and suggest the notion of regime of attention as a focus for future biosocial research on urban mental health.
Background Urinary tract infection (UTI) is one of the most common bacterial infections responsible for increased annual incidence of antimicrobial resistance (AMR) cases. Clinical diagnosis of UTI AMR relies heavily on conventional urine culture and antibiotic susceptibility testing (AST) which has a turnaround time of ∼3 days. Often, irrespective of the infection status, antibiotics are prescribed to patients even before the test results are available, leading to non-judicious use of antibiotics. Over the years, several technologies have been developed for the rapid detection and diagnosis of UTI AMR, however, most of them are limited to traditional microbiological techniques and large laboratory equipment that are not readily available in low-to-middle income countries (LMICs). To address these diagnostic limitations, we are developing a rapid and affordable UTI-AMR diagnostic microfluidic device that is clinical friendly aimed at improving UTI management and AMR stewardship. Results Our device enables the flow of a large volume of urine specimens for the capture/enrichment of uropathogenic bacteria and determination of AST via a porous membrane that is augmented with a multifunctional polymer-based material. Important objectives for the development of UTI AMR diagnostic microfluidic device are: (i) development of a multifunctional polymer-based material; and (ii) validation of UTI AMR diagnostic device. We have successfully developed a polysaccharide-based platform to (i) selectively capture uropathogenic bacteria from urine specimen by immobilizing concanavalin A (con A) lectin as bacterial capture agent on the polymer surface via chemical modification; (ii) encapsulate and release bacterial nutrient media and antibiotics for AST; and (iii) detect AST via encapsulation of bacterial growth indicator. In addition, we have also determined the development of methacrylate-based and acrylamide-based synthetic polymer-based material for our application. Further, we have demonstrated the uniform augmentation of the polysaccharide-based polymer onto porous membrane via dip-coating technique for on-chip bacterial capture/enrichment and AST in fluid (urine) flow conditions. The porous membrane is a conducting material which enables us to perform electrochemical measurements such as impedance spectroscopy that accelerates the detection process of antibiotic susceptibility. As a proof-of-concept, we have determined the capture of biosafety level I Escherichia coli expressing kanamycin resistance gene on chemically surface modified polysaccharide-based polymer containing con A and the antibiotic susceptibility of captured bacteria against different antibiotics with and without the porous membrane. We have quantitatively determined the limit of detection of E. coli on multifunctional polysaccharide-based polymer material. Conclusions The utility of the UTI AMR microfluidic device in clinical settings enables clinicians to make informed decisions on the most appropriate antibiotic for treatment in less than a day. Integration of impedance spectroscopy will further accelerate the detection by significantly reducing the time of detection. Further, the device allows for off-chip analysis by retrieving the captured uropathogenic bacteria to perform high throughput sequencing for identifying AMR genetic determinants. Therefore, with the ability to selectively capture uropathogenic bacteria and determine AST in a short time, our technology has the potential to overcome some of the current limitations in UTI AMR diagnostics.
Conditionally automated cars share the driving task with the driver. When the control switches from one to another, accidents can occur, especially when the car emits a takeover request (TOR) to warn the driver that they must take the control back immediately. The driver’s physiological state prior to the TOR may impact takeover performance and as such was extensively studied experimentally. However, little was done about using Machine Learning (ML) to cluster natural states of the driver. In this study, four unsupervised ML algorithms were trained and optimized using a dataset collected in a driving simulator. Their performances for generating clusters of physiological states prior to takeover were compared. Some algorithms provide interesting insights regarding the number of clusters, but most of the results were not statistically significant. As such, we advise researchers to focus on supervised ML using ground truth labels after experimental manipulation of drivers’ states.
In the 1980s, Switzerland’s Jura Arc region was a globally competitive ‘new industrial space’ in the Third Industrial Revolution’s flexible accumulation regime based on information and communication technology (ICT) and automation processes. Recently, this nowadays ‘old industrial space’ has been experiencing the implementation of Industry 4.0. Caught between the use of existing productive assets and the development of platform-based market ecosystems, this region illustrates the challenges inherent in implementing ‘forking innovation’, which requires the development not only of new business models, but also of collaborative and investment models in order to scale up and increase local value capture.
This article demonstrates person localization using a hybrid system consisting of an electromagnetic positioning system and a depth camera to authorize access control. The ultimate aim of this system is to distinguish moving people in a defined area by tracking the RF device and the people. It focuses on the application and incorporation of the received data from these two systems. Both systems send data simultaneously which is stored in a Docker container for further analysis. The data is processed in real-time to track the movement of the targets. The centralized database monitoring grants secure access to the information. The motive for using this hybrid system lies in the ever-growing need for accurate position determination for indoor and complex environments. Track and tracing are especially important in access-control applications. The system has a great impact on real-life access-control applications in malls, shops, train stations, and generally everyplace where the access control requires monitoring. The non-blocking feature plus the accuracy can provide ease of use for the users. Moreover, employing a low-frequency tag system does not suffer from the multipath effect and non-line of sight problems that are inevitable for indoor applications. By extending the number of users for a larger area, this system can replace traditional security gates with a pleasant look and comfortable application.
The problem of stress detection and classification has attracted a lot of attention in the past decade. It has been tackled with mainly two different approaches, where signals were either collected in ambulatory settings, which can be limited to the period of presence in the hospital, or in continuous mode in the field. A sensor-based continuous measurement of stress in daily life has a potential to increase awareness of patterns of stress occurrence. In this work, we first present a data-flow infrastructure suitable for two types of studies that conforms with the data protection requirements of the ethics committee monitoring the research on humans. The detection and binary classification of stress events is compared with three different machine learning models based on the features (meta-data) extracted from physiological signals acquired in laboratory conditions and ground-truth stress level information provided by the subjects themselves via questionnaires associated with these features. The main signals considered in current classification are electro-dermal activity (EDA) and blood volume pulse (BVP) signals. Different models are compared and the best configuration yields an \(F_1\) score of 0.71 (random baseline: 0.48). The importance on prediction of phasic and tonic EDA components is also investigated. Our results also pave the way for further work on this topic with both machine learning approaches and signal processing directions.
The construction of large microbial fuel cells (MFCs) and their long-term reliability are current challenges. MFCs generate power while purifying wastewater, save electricity, avoid pollutant stripping into the air and are a source of CO2. To understand larger MFCs, a 1000-L MFC was designed. It was built from transparent polyester and electrodes were from reticulated vitreous carbon. Four power management devices were connected to an ensemble of 64 MFC units and assembled as a 12 m long MFC. Two Raspberry and a personal computer with Python programmed software automatized power management. The MFC was run for one year under maximum power point tracking (MPPT). Temperatures between 11.5 °C and 21 °C corresponded to WWTP conditions. The reactor shared electrolytes within 12 m long half-cells and 80–95% COD was removed generating 0.015 to 0.060 kWh/m³ with an energy efficiency of 5.8–12.1%. Voltage reversal were seen as potential imbalances among MFC units and all self-healing. Ammonium removal reached 48%, phosphorous was reduced to 0.59 mg/L, and micropollutants degraded by 67%. Biofilm mapping by 16S rRNA metagenomics indicated bi-sectorial metabolic properties. 10 Major genera were essential in the elongated scale up MFC generating electricity, reduced energy needed, and purified wastewater.
Takeover requests in conditionally automated vehicles are a critical point in time that can lead to accidents, and as such should be transmitted with care. Currently, several studies have shown the impact of using different modalities for different psychophysiological states, but no model exists to predict the takeover quality depending on the psychophysiological state of the driver and takeover request modalities. In this paper, we propose a machine learning model able to predict the maximum steering wheel angle and the reaction time of the driver, two takeover quality metrics. Our model is able to achieve a gain of 42.26% on the reaction time and 8.92% on the maximum steering wheel angle compared to our baseline. This was achieved using up to 150 s of psychophysiological data prior to the takeover. Impacts of using such a model to choose takeover modalities instead of using standard takeover requests should be investigated.
The spread of deep learning on embedded devices has prompted the development of numerous methods to optimise the deployment of deep neural networks (DNN). Works have mainly focused on: i) efficient DNN architectures, ii) network optimisation techniques such as pruning and quantisation, iii) optimised algorithms to speed up the execution of the most computational intensive layers and, iv) dedicated hardware to accelerate the data flow and computation. However, there is a lack of research on cross-level optimisation as the space of approaches becomes too large to test and obtain a globally optimised solution. Thus, leading to suboptimal deployment in terms of latency, accuracy, and memory. In this work, we first detail and analyse the methods to improve the deployment of DNNs across the different levels of software optimisation. Building on this knowledge, we present an automated exploration framework to ease the deployment of DNNs. The framework relies on a Reinforcement Learning search that, combined with a deep learning inference framework, automatically explores the design space and learns an optimised solution that speeds up the performance and reduces the memory on embedded CPU platforms. Thus, we present a set of results for state-of-the-art DNNs on a range of Arm Cortex-A CPU platforms achieving up to 4× improvement in performance and over 2× reduction in memory with negligible loss in accuracy with respect to the BLAS floating-point implementation.
In conditionally automated driving, drivers do not have to constantly monitor their vehicle but they must be able to take over control when necessary. In this paper, we assess the impact of instructions about limitations of automation and the presentation of context-related information through a mobile application on the situation awareness and takeover performance of drivers. We conducted an experiment with 80 participants in a fixed-base driving simulator. Participants drove for an hour in conditional automation while performing secondary tasks on a tablet. Besides, they had to react to five different takeover requests. In addition to the assessment of behavioral data (e.g. quality of takeover), participants rated their situation awareness after each takeover situation. Instructions and context-related information on limitations combined showed encouraging results to raise awareness and improve takeover performance.
Next generation of embedded Information and Communication Technology (ICT) systems are interconnected and collaborative systems able to perform autonomous tasks. The remarkable expansion of the embedded ICT market, together with the rise and breakthroughs of Artificial Intelligence (AI), have put the focus on the Edge as it stands as one of the keys for the next technological revolution: the seamless integration of AI in our daily life. However, training and deployment of custom AI solutions on embedded devices require a fine-grained integration of data, algorithms, and tools to achieve high accuracy and overcome functional and non-functional requirements. Such integration requires a high level of expertise that becomes a real bottleneck for small and medium enterprises wanting to deploy AI solutions on the Edge , which, ultimately, slows down the adoption of AI on applications in our daily life. In this work, we present a modular AI pipeline as an integrating framework to bring data, algorithms, and deployment tools together. By removing the integration barriers and lowering the required expertise, we can interconnect the different stages of particular tools and provide a modular end-to-end development of AI products for embedded devices. Our AI pipeline consists of four modular main steps: (i) data ingestion, (ii) model training, (iii) deployment optimization, and (iv) the IoT hub integration. To show the effectiveness of our pipeline, we provide examples of different AI applications during each of the steps. Besides, we integrate our deployment framework, Low-Power Deep Neural Network (LPDNN), into the AI pipeline and present its lightweight architecture and deployment capabilities for embedded devices. Finally, we demonstrate the results of the AI pipeline by showing the deployment of several AI applications such as keyword spotting, image classification, and object detection on a set of well-known embedded platforms, where LPDNN consistently outperforms all other popular deployment frameworks.
Based on a case study and expert interviews, the present paper takes a detailed look at the origin, mechanisms and results of the Speedy Tuesday operations conducted by the Omega watch brand and the Speedy Tuesday community on Instagram. Our findings point to a profound evolution in communication codes that are based on a co-creation dynamic with lovers of the brand, relayed on social networks. This development has led to what we have called “communication-distribution morphing”, in other words, a hybridization of communication and distribution channels.
Autonomous vehicles are developing rapidly and will lead to a significant change in the driver's role: he/she will have to move from the role of actor to the role of supervisor. Indeed, the driver will soon be able to perform a secondary task but he/she must be able to take over control in the event of a critical situation that is not managed by the autonomous system. This implies that the role of new interfaces and interactions within the vehicle is important to take into account. This article describes the design of an application that provides the driver with information about the environment perceived by his/her vehicle in the form of modules. This application is displayed as split screen on a tablet by which a secondary task can be performed. Initial tests were carried out with this application in a driving simulator. They made it possible to test the acceptance of the application and the clarity of the information transmitted. The results generally showed that the participants correctly identified some of the factors limiting the proper functioning of the autonomous pilot while performing a secondary task on a tablet.
Internet of Things for medical devices is revolutionizing healthcare industry by providing platforms for data collection via cloud gateways and analytic. In this paper, we propose a process for developing a proof of concept solution for sleep detection by observing a set of ambulatory physiological parameters in a completely non-invasive manner. Observing and detecting the state of sleep and also its quality, in an objective way, has been a challenging problem that impacts many medical fields. With the solution presented here, we propose to collect physiological signals from wearable devices, which in our case consist of a smart wristband equipped with sensors and a protocol for communication with a mobile device. With machine learning based algorithms, that we developed, we are able to detect sleep from wakefulness in up to 93% of cases. The results from our study are promising with a potential for novel insights and effective methods to manage sleep disturbances and improve sleep quality.
Autonomous vehicles are developing rapidly and will lead to a significant change in the driver's role: s/he will have to move from the role of actor to the role of supervisor. Indeed, s/he will soon be able to perform a secondary task but s/he must be able to take over control when a critical situation is not managed by the driving system. The role of new interfaces and interactions within the vehicle is important to take into account. This article describes the design of an application that provides the driver with information about the environment perceived by the vehicle. This application is displayed as split screen on a tablet by which a secondary task can be performed. The results of initial experiment showed that the participants correctly identified all the factors limiting the proper functioning of the driving system while performing a secondary task on the tablet.
Driver distraction is a major issue in manual driving, causing more than 30'000 fatal crashes on US roadways in 2015 only . As such, it is widely studied in order to increase driving safety. Many studies show how to detect driver distraction using Machine Learning algorithms and driver psychophysiological data. In this study, we investigate the trade-off between efficiency and privacy while predicting driver distraction. Specifically, we want to assess the impact on the estimation of the driver state without access to his/her psychophysiological data. Different Machine Learning models (Convolutional Neural Networks, K-NN and Random forest) are implemented to evaluate the validity of the distraction detection with and without access to psychophysiological data. The results show that a Convolutional Neural Network model is still able to detect driver distraction without access to psychophysiological features, with an f1-score of 97.11%, losing only 1.37% in the process.
In the context of highly automated driving, the driver has to be aware of driving risks and to take over control of the car in hazardous situations. The goal of this paper is to categorize and analyze the factors that lead to such critical scenarios. To this purpose, we analyzed limitations of Advanced Driver-Assistance Systems (ADAS) extracted from owner manuals of 12 partially automated cars available on the market. A taxonomy with 6 macro-categories and 26 micro-categories is proposed to classify and better understand the limitations of these vehicles. We also investigated if these limitations are conveyed to the driver through Human-Machine Interaction (HMI) in the car. Some suggestions are made to better communicate these limitations to the driver in order to raise his/her situation awareness.
With the increasing use of automation, users tend to delegate more tasks to the machines. Such complex systems are usually developed with "black box" Artificial Intelligence (AI), which makes these systems difficult to understand for the user. This assumption is particularly true in the field of automated driving since the level of automation is constantly increasing via the use of state-of-the-art AI solutions. We believe it is important to investigate the field of Explainable AI (XAI) in the context of automated driving since interpretability and transparency are key factors for increasing trust and security. In this workshop, we aim at gathering researchers and industry practitioners from different fields to brainstorm about XAI with a special focus on human-vehicle interaction. Questions like "what kind of explanation do we need", "which is the best trade-off between performance and explainability" and "how granular should the explanations be" will be addressed in this workshop.
A fully parameterized three-dimensional model with specific dimensions has been developed in ANSYS for an insulin injection pen used by diabetic persons. The insulin injection pen has a smart pen cap which hosts four electrodes used for the smart pen cap electrode capacitive measurement. The addition of the smart cap on top of the insulin injection pens is novel and essential for storing and transmitting injection dose and time data to help patients successfully manage their treatment. The simulations can be used to decide the number and exact shape of the electrodes, as well as to evaluate different misalignment and asymmetries of the electrode fabrication process or of the liquid misplacement. Using Maxwell 3D tool the electrode capacitance was numerically evaluated, which is necessary for sensing and ultimately for insulin dose precise detection. Experimental results have been provided using an AD7746 high-resolution sigma-delta capacitance-to-digital converter (CDC). Simulation and experimental results for the sense electrode capacitance, in the case of both smart pen cap and complete insulin injection pen + smart pen cap system, have been obtained, using two different configurations (1 vs 3 and 2 vs 2 respectively). Smart pen cap electrode capacitance variation for different insulin fill states has been numerically evaluated and the linear behaviour of the injection has been proven.
Institution pages aggregate content on ResearchGate related to an institution. The members listed on this page have self-identified as being affiliated with this institution. Publications listed on this page were identified by our algorithms as relating to this institution. This page was not created or approved by the institution. If you represent an institution and have questions about these pages or wish to report inaccurate content, you can contact us here.