With the Internet of Things paradigm, the data generated by the rapidly increasing number of connected devices lead to new possibilities, such as using machine learning for activity recognition in smart environments. However, it also introduces several challenges. The sensors of different devices might be mobile and of different types, i.e. there is a need to handle streaming data from a dynamic and heterogeneous set of sensors. In machine learning, the performance is often linked to the availability and quality of annotated data. Annotating data is in general costly, but it can be even more challenging if there is not any, or a very small amount of, annotated data to train the model on at the start of learning. To handle these issues, we implement interactive and adaptive systems. By including human-in-the-loop, which we refer to as interactive machine learning, the input from users can be utilized to build the model. The type of input used in interactive machine learning is typically annotations of the data, i.e. correctly labelled data points. Generally, it is assumed that the user always provides correct labels in accordance with the chosen interactive learning strategy. In many real-world applications these assumptions are not realistic however, as users might provide incorrect labels or not provide labels at all in line with the chosen strategy.

In this thesis we explore which interactive learning strategy types are possible in the given scenario and how they affect performance, as well as the effect of machine learning algorithms on the performance. We also study how a user who is not always reliable, i.e. who does not always provide a correct label when expected to, can affect performance. We propose a taxonomy of interactive online machine learning strategies and test how the different strategies affect performance through experiments on multiple datasets. Simulated experiments are compared to experiments with human participants, to verify the results. The findings show that the overall best performing interactive learning strategy is one where the user provides labels when current estimations are incorrect, but that the best performing machine learning algorithm depends on the problem scenario. The experiments also show that a decreased reliability of the user leads to decreased performance, especially when there is a limited amount of labelled data. The robustness of the machine learning algorithms differs, where e.g. Naïve Bayes classifier is better at handling a lower reliability of the user. We also present a systematic literature review on machine teaching, a subfield of interactive machine learning where the human is proactive in the interaction. The study shows that the area of machine teaching is rapidly evolving with an increased number of publications in recent years. However, as it is still maturing, there exists several open challenges that would benefit from further exploration, e.g. how human factors can affect performance.

Although the availability of sensor data is becoming prevalent across many domains, it still remains a challenge to make sense of the sensor data in an efficient and effective manner in order to provide users with relevant services. The concept of virtual sensors provides a step towards this goal, however they are often used to denote homogeneous types of data, generally retrieved from a predetermined group of sensors. The DIVS (Dynamic Intelligent Virtual Sensors) concept was introduced in previous work to extend and generalize the notion of a virtual sensor to a dynamic setting with heterogenous sensors. This paper introduces a refined version of the DIVS concept by integrating an interactive machine learning mechanism, which enables the system to take input from both the user and the physical world. The paper empirically validates some of the properties of the DIVS concept. In particular, we are concerned with the distribution of different budget allocations for labelled data, as well as proactive labelling user strategies. We report on results suggesting that a relatively good accuracy can be achieved despite a limited budget in an environment with dynamic sensor availability, while proactive labeling ensures further improvements in performance.

In interactive machine learning, human users and learning algorithms work together in order to solve challenging learning problems, e.g. with limited or no annotated data or trust issues. As annotating data can be costly, it is important to minimize the amount of annotated data needed for training while still getting a high classification accuracy. This is done by attempting to select the most informative data instances for training, where the amount of instances is limited by a labelling budget. In an online learning setting, the decision of whether or not to select an instance for labelling has to be done on-the-fly, as the data arrives in a sequential order and is only valid for a limited time period. We present a taxonomy of interactive online machine learning strategies. An interactive learning strategy determines which instances to label in an unlabelled dataset. In the taxonomy we differentiate between interactive learning strategies when the computer controls the learning process (active learning) and those when human users control the learning process (machine teaching). We then make a distinction between what triggers the learning: active learning could be triggered by uncertainty, time, or randomly, whereas machine teaching could be triggered by errors, state changes, time, or factors related to the user. We also illustrate the taxonomy by implementing versions of the different strategies and performing experiments on a benchmark dataset as well as on a synthetically generated dataset. The results show that the choice of interactive learning strategy affects performance, especially in the beginning of the online learning process, when there is a limited amount of labelled data.

Interactive Machine Learning (ML) has the potential to lower the manual labelling effort needed, as well as increase classification performance by incorporating a human-in-the loop component. However, the assumptions made regarding the interactive behaviour of the human in experiments are often not realistic. Active learning typically treats the human as a passive, but always correct, participant. Machine teaching provides a more proactive role for the human, but generally assumes that the human is constantly monitoring the learning process. In this paper, we present an interactive online framework and perform experiments to compare active learning, machine teaching and combined approaches. We study not only the classification performance, but also the effort (to label samples) and attention (to monitor the ML system) required of the human. Results from experiments show that a combined approach generally performs better with less effort compared to active learning and machine teaching. With regards to attention, the best performing strategy varied depending on the problem setup.

The advances in Internet of things lead to an increased number of devices generating and streaming data. These devices can be useful data sources for activity recognition by using machine learning. However, the set of available sensors may vary over time, e.g. due to mobility of the sensors and technical failures. Since the machine learning model uses the data streams from the sensors as input, it must be able to handle a varying number of input variables, i.e. that the feature space might change over time. Moreover, the labelled data necessary for the training is often costly to acquire. In active learning, the model is given a budget for requesting labels from an oracle, and aims to maximize accuracy by careful selection of what data instances to label. It is generally assumed that the role of the oracle only is to respond to queries and that it will always do so. In many real-world scenarios however, the oracle is a human user and the assumptions are simplifications that might not give a proper depiction of the setting. In this work we investigate different interactive machine learning strategies, out of which active learning is one, which explore the effects of an oracle that can be more proactive and factors that might influence a user to provide or withhold labels. We implement five interactive machine learning strategies as well as hybrid versions of them and evaluate them on two datasets. The results show that a more proactive user can improve the performance, especially when the user is influenced by the accuracy of earlier predictions. The experiments also highlight challenges related to evaluating performance when the set of classes is changing over time.

Interactive machine learning (ML) adds a human-in-the-loop aspect to a ML system. Even though the input from human users to the system is a central part of the concept, the uncertainty caused by the human feedback is often not considered in interactive ML. The assumption that the human user is expected to always provide correct feedback, typically does not hold in real-world scenarios. This is especially important for when the cognitive workload of the human is high, for instance in online learning from streaming data where there are time constraints for providing the feedback. We present experiments of interactive online ML with human participants, and compare the results to simulated experiments where humans are always correct. We found combining the two interactive learning paradigms, active learning and machine teaching, resulted in better performance compared to machine teaching alone. The results also showed an increased discrepancy between the experiments with human participants and the simulated experiments when the cognitive workload was increased. The findings suggest the importance of taking uncertainty caused by human factors into consideration in interactive ML, especially in situations which requires a high cognitive workload for the human.

The advances in Internet of Things lead to an increased number of devices generating and streaming data. These devices can be useful data sources for Activity Recognition by using Machine Learning. However, as the set of available sensors may vary over time, e.g. due to mobility of the sensors and technical failures, the feature space might also change over time. Moreover, the labelled data necessary for the training is often costly to acquire. Active Learning is a type of Interactive Machine Learning where the model is given a budget for requesting labels from an oracle, and aims to maximize accuracy by careful selection of what data points to label. It is generally assumed that a query always gets a correct response, but in many real-world scenarios this is not a realistic assumption. In this work we investigate different Proactive Learning strategies, which explore the human factors of the oracle and aspects that might influence a user to provide or withhold labels. We implemented four proactive strategies and hybrid versions of them. They were evaluated on two datasets to examine how a more proactive, or reluctant, user affects performance. The results show that a more proactive user can improve the performance, especially when the user is influenced by the accuracy of earlier predictions. The experiments also highlight challenges related to evaluating performance when the set of classes is changing over time.

In interactive machine learning it is important to select the most informative data instances to label in order to minimize the effort of the human user. There are basically two categories of interactive machine learning. In the first category, active learning, it is the computational learner that selects which data to be labelled by the human user, whereas in the second one, machine teaching, the selection is done by the human teacher. It is often assumed that the human user is a perfect oracle, i.e., a label will always be provided in accordance with the interactive learning strategy and that this label will always be correct. In real-world scenarios however, these assumptions typically do not hold. In this work, we investigate how the reliability of the user providing labels affects the performance of online machine learning. Specifically, we study reluctance, i.e., to what extent the user does not provide labels in accordance with the strategy, and fallibility, i.e., to what extent the provided labels are incorrect. We show results of experiments on a benchmark dataset as well as a synthetically created dataset. By varying the degree of reluctance and fallibility of the user, the robustness of the different interactive learning strategies and machine learning algorithms is explored. The experiments show that there is a varying robustness of the strategies and algorithms. Moreover, certain machine learning algorithms are more robust towards reluctance compared to fallibility, while the opposite is true for others

With advances in Internet of Things many opportunities arise if the challenges of continual learning in a multimodal setting can be tackled. One common issue in Online Learning is to obtain labelled data, as this generally is costly. Active Learning is a popular approach to collect labelled data efficiently, but in general includes unrealistic assumptions. In this work we present a first step towards a taxonomy of Interactive Learning strategies in a multimodal and dynamic setting. By relaxing assumptions of standard Active Learning, the strategies become better suited for real-world settings and can achieve better performance.

The impact of Twitter can be strongly felt across a number of areas of professional and day-to-day life. By combining quantitative and qualitative methods, this article analyzes how young college students create content on Twitter in an academic setting, and how they perceive this content creation. To this end, a purposive sample of three groups of students was formed, a sample of 10,291 published tweets was collected, and a focus group was set up for each group of students. Among other findings, this study shows that students prioritize academic instructions and requirements and report that the way they use the platform depends on whether it is for an academic assignment or not.

This paper explores the problem of determining the time of an analogue wristwatch by developing two systems and conducting a comparative study. The first system uses OpenCV to find the watch hands and applies geometrical techniques to calculate the time. The second system uses Machine Learning by building a neural network to classify images in Tensorflow using a multi-labelling approach. The results show that in a set environment the geometric-based approach performs better than the Machine Learning model. The geometric system predicted time correctly with an accuracy of 80% whereas the best Machine Learning model only achieves 74%. Experiments show that the accuracy of the neural network model did increase when using data augmentation, however there was no significant improvement when adding synthetic data to our training set.

The primary objective of intelligent transport systems (ITS) is to improve road safety and efficiency in the transport sector and are addressed in Internet of vehicles (IoV) based solutions, peer to peer vehicle data sharing, inter vehicle and vehicle to infrastructure communication channels. Effectiveness of an IoV solution is dependent on the robustness of the wireless communication technology. Performance of ZigBee, Wi-Fi and DSRC communication technologies for deployment in an IoV system is investigated by performing simulations in an identified platform. It was found that ZigBee, Wi-Fi and DSRC can offer successful exchange of data with failure rate of less than 1% for low frequency of communication events while Wi-Fi and DSRC can offer this performance at even higher frequencies of exchange events. Findings of the research will be used to design and test road safety and traffic management mechanisms for an IoV system.

We describe Bar Chart Ball, a game where players indirectly control a ball by modifying a bar chart that the ball rests on. The bar chart displays real-world demographic data about the UK, and the player modifies the chart by selecting which aspect of the data to focus on. By making data selection a core game mechanic, in fact the only game mechanic, we advance a novel and simple way of building game content from data, and of making data visualisation playable.

With the growth of procedural content generation in game development, there is a need for a viable generative method to give context and make sense of the content within game space. We propose procedural narrative as context through objectives, as a useful means to structure content in games. In this paper, we present and describe an artifact developed as a sub-system to the Evolutionary Dungeon Designer (EDD) that procedurally generates objectives for the dungeons created with the tool. The quality of the content within rooms is used to generate objectives, and together with the distributions and design of the dungeon, main and side objectives are formed to maximize the usage of game space and create a proper context.  

Monitoring asthma is essential for self-management. However, traditional monitoring methods require high levels of active engagement, and some patients may find this tedious. Passive monitoring with mobile-health devices, especially when combined with machine-learning, provides an avenue to reduce management burden. Data for developing machine-learning algorithms are scarce, and gathering new data is expensive. A few datasets, such as the Asthma Mobile Health Study, are publicly available, but they only consist of self-reported diaries and lack any objective and passively collected data. To fill this gap, we carried out a 2-phase, 7-month AAMOS-00 observational study to monitor asthma using three smart-monitoring devices (smart-peak-flow-meter/smart-inhaler/smartwatch), and daily symptom questionnaires. Combined with localised weather, pollen, and air-quality reports, we collected a rich longitudinal dataset to explore the feasibility of passive monitoring and asthma attack prediction. This valuable anonymised dataset for phase-2 of the study (device monitoring) has been made publicly available. Between June-2021 and June-2022, in the midst of UK's COVID-19 lockdowns, 22 participants across the UK provided 2,054 unique patient-days of data.

Asthma monitoring is an important aspect of patient self-management. However, due to its repetitive nature, patients can find long-term monitoring tedious. Mobile health can provide an avenue to monitor asthma without needing high levels of active engagement, and instead rely on passive monitoring. In our recent AAMOS-00 study, we collected mobile health data over six months from 22 asthma patients using passive and active monitoring technology, including smartwatch, peak flow measurements, and daily asthma diaries.

Compliance to smartwatch monitoring was found to lie between the compliance to complete daily asthma diaries and measuring daily peak flow. However, some study participants faced technical issues with the devices which could have affected the relative compliance of the monitoring tasks.

Moreover, as evidenced by standard usability questionnaires, we found that the AAMOS-00 study’s data collection system was similar in quality to other studies and published apps.

INTRODUCTION: Supported self-management empowering people with asthma to detect early deterioration and take timely action reduces the risk of asthma attacks. Smartphones and smart monitoring devices coupled with machine learning could enhance self-management by predicting asthma attacks and providing tailored feedback.We aim to develop and assess the feasibility of an asthma attack predictor system based on data collected from a range of smart devices.

METHODS AND ANALYSIS: A two-phase, 7-month observational study to collect data about asthma status using three smart monitoring devices, and daily symptom questionnaires. We will recruit up to 100 people via social media and from a severe asthma clinic, who are at risk of attacks and who use a pressurised metered dose relief inhaler (that fits the smart inhaler device).Following a preliminary month of daily symptom questionnaires, 30 participants able to comply with regular monitoring will complete 6 months of using smart devices (smart peak flow meter, smart inhaler and smartwatch) and daily questionnaires to monitor asthma status. The feasibility of this monitoring will be measured by the percentage of task completion. The occurrence of asthma attacks (definition: American Thoracic Society/European Respiratory Society Task Force 2009) will be detected by self-reported use (or increased use) of oral corticosteroids. Monitoring data will be analysed to identify predictors of asthma attacks. At the end of the monitoring, we will assess users' perspectives on acceptability and utility of the system with an exit questionnaire.

ETHICS AND DISSEMINATION: Ethics approval was provided by the East of England - Cambridge Central Research Ethics Committee. IRAS project ID: 285 505 with governance approval from ACCORD (Academic and Clinical Central Office for Research and Development), project number: AC20145. The study sponsor is ACCORD, the University of Edinburgh.Results will be reported through peer-reviewed publications, abstracts and conference posters. Public dissemination will be centred around blogs and social media from the Asthma UK network and shared with study participants.

In this paper, we highlight how including technology, movement or play can boost a design process but with unbalanced amounts can also hamper the process. We provide a set of examples where we miscalculated the amount of technology, movement, or play that was needed in a design activity in such a way that it became counterproductive and for each example mention possible adaptations. Finally, we highlight three existing approaches that can balance the overabundance of technology, movement, and play in design processes: activity-centered design, somaesthetic design, and perspective-changing movement-based design.  

Objectives: Digital transformation (DT) has become an imperative within research and practice. Still, companies experience obstacles when trying to pursue a successful DT. Numerous scientific sources have dealt with the identification of barriers to DT. In doing so, scientists have already produced reviews to identify and classify the barriers to DT. However, the scientific work often relates to specific company contexts. In addition, there is no structured overview of the literature on how to overcome barriers to DT. The mere identification provides an incomplete view on the barriers to DT and needs to be complemented by approaches to overcome them. Thus, our research question is: Which barriers and recommendations for action to DT exist within the scientific literature and how can they be clustered according to a holistic sociotechnical perspective?

Data and Method: Our study follows the approach of a structured literature review combined with additional focus group work to generate a concept matrix to structure barriers and recommendations for action. The conducted literature search generated 562 articles (without duplicates). After a first screening 148 articles were deemed to be applicable for our study. A more in-depth qualitative check generated 99 relevant articles. Different sections of these articles were openly coded into 178 barriers and 161 recommendations for action. These codes were then clustered in focus group sessions.

Results: The result of our research approach is a framework containing clustered barriers and cluster-related recommendations for overcoming. The following clusters were identified: individual, technical, financial, organizational alignment, organizational design, organizational culture, market environment, and regulatory. Our review discloses that not all clusters receive equal attention in the literature. In particular, organizational culture is given less consideration, while especially individual, technical and financial is in focus. The identified recommendations for action show that not all barriers can be solved by the companies themselves but require governmental support instead.

Conclusions: Our study generated a holistic framework. As barriers either slow down or even entirely hinder DT, understanding their nature is essential. Our discussion reveals that several barriers are contrasting each other. This implies that managers need to carefully balance DT initiatives. The framework provides guidance on doing so. The findings also provide a solid foundation for future research, as our literature review presents a state-of-the-art of current research and reveals research gaps.

Today, the Internet of Things (IoT) is mainly associated with vertically integrated systems that often are closed and fragmented in their applicability. To build a better IoT ecosystem, the open IoT platform has become a popular term in the recent years. However, this term is usually used in an intuitive way without clarifying the openness aspects of the platforms. The goal of this paper is to characterize the openness types of IoT platforms and investigate what makes them open. We conducted a systematic mapping study by retrieving data from 718 papers. As a result of applying the inclusion and exclusion criteria, 221 papers were selected for review. We discovered 46 IoT platforms that have been characterized as open, whereas 25 platforms are referred as open by some studies rather than the platforms themselves. We found that the most widely accepted and used open IoT platforms are NodeMCU and ThingSpeak that together hold a share of more than 70% of the declared open IoT platforms in the selected papers. The openness of an IoT platform is interpreted into different openness types. Our study results show that the most common openness type encountered in open IoT platforms is open-source, but also open standards, open APIs, open data and open layers are used in the literature. Finally, we propose a new perspective on how to define openness in the context of IoT platforms by providing several insights from the different stakeholder viewpoints.

While security is often recognized as a top priority for organizations and a push for competitive advantage, repeatedly, Internet of Things (IoT) products have become a target of diverse security attacks. Thus, orchestrating smart services and devices in a more open, standardized and secure way in IoT environments is yet a desire as much as it is a challenge. In this paper, we propose a model for IoT practitioners and researchers, who can adopt a sound security thinking in parallel with open IoT technological developments. We present the state-of-the-art and an empirical study with IoT practitioners. These efforts have resulted in identifying a set of openness and security thinking criteria that are important to consider from an IoT ecosystem point of view. Openness in terms of open standards, data, APIs, processes, open source and open architectures (flexibility, customizability and extensibility aspects), by presenting security thinking tackled from a three-dimensional point of view (awareness, assessment and challenges) that highlight the need to develop an IoT security mindset. A novel model is conceptualized with those characteristics followed by several key aspects important to design and secure future IoT systems.

In this paper we propose a model for Internet of Things (IoT) practitioners and researchers on how to use security thinking in parallel with the IoT technological developments. While security is recognized as a top priority, repeatedly, IoT products have become a target by diverse security attacks. This raises the importance for an IoT security mindset that contributes to building more holistic security measures. In understanding this, we present the state-of-the-art in IoT security. This resulted in the identification of three dimensions (awareness, assessment and challenges) that are needed to develop an IoT security mindset. We then interviewed four security and IoT-related experts from three different organizations that formed the basis for our pilot study to test the model. Our results show that the identified three-dimensional model highlights continuous security thinking as a serious matter to sustain IoT development with positive outcomes for its users.

The traditional way of teaching may no longer be sufficient to cope with current requirements specifically in the Internet of Things (IoT) domain. The case for this paper is related to an introductory programming course on JavaScript for the period of 2016-2018. In this study a multi-method approach for data collection is utilized. Project-Based Learning (PBL), Scrum and rapid prototyping are utilized to support student projects over the three years. Students developed a number of prototypes for various IoT domains related to ongoing research projects within our research center. The results show that students could easily use their JavaScript knowledge for any type of IoT development. PBL, Scrum and rapid prototyping help addressing uncertainties during the projects and balancing the team efforts for learning, development, problem solving and creativity. One of the outcomes of this paper confirms that smaller team sizes of students perform better during the project lifetime. In conclusion, focusing on knowledge increase, teamwork, collaboration, interaction, constant feedback, and adaptability should be considered a priority while exploiting teaching approaches such as PBL, Scrum and rapid prototyping for IoT development.

Fast growth in a number of connected devices, heterogeneity, constrained resources, privacy, software upgrades and operational environment create important security related challenges in Internet of Things (IoT) domain. In this research, a literature survey of the state of the art is conducted related to security aspects. The results are validated by conducting qualitative interviews with IoT practitioners. The efforts have resulted towards identifying several security trends and challenges and some design aspects to be considered by IoT practitioners. The outcomes point towards that security is not only a technical problem but is more of an awareness, mindset, people and process issue. In this paper, a novel model is conceptualized with openness and security characteristics followed with several key aspects important to design an IoT system. This model emphasizes the human in the loop aspect, which would help to determine the dynamic requirements and design principles of IoT systems in a more open, common and secure way.

With the rise of digital technologies, digital transformation (DT) has become an issue in the field of higher education. In higher education institutions and enterprises alike, DT means to digitalize internal processes and offer digital services and products. There are barriers that must be overcome to master this challenge. Our study follows an explorative mixed methods design. We identify the barriers to DT and transfer them to a research model. We examine the influence of individually perceived barriers on the DT process, thus contributing to the theoretical foundation of DT barriers in higher education institutions. This paper offers an approved scale to measure barriers to DT and a valid operationalization of DT barriers. The identified predictors can explain over 50% of the alteration problems of the DT process. Results indicate that management students have a significantly low level of concern regarding their privacy and traceability but demand more commitment at the organizational level.

Companies expect significant long-term gains in efficiency and productivity through digital transformation (DT). New ways of combining products, processes, and data-driven services, as well as new business models emerge. However, the rapid development of the DT leads to constraints regarding its realization. Barriers hinder companies to realize possible advantages out of DT. If firms promptly recognize potential barriers, they can reflect upon these challenges and can take well-coordinated countermeasures. Social, technical and socio-technical problems address different stakeholders and ask for specific solutions. Therefore, our study aims at developing a taxonomy for barriers to DT to enable researchers and practitioners to identify and classify existing barriers. For deriving the dimensions and characteristics, we collected data by conducting 46 semi-structured interviews with experts and enriched these by looking at the literature on DT barriers.

Digital Transformation (DT) is expected to have a massive impact on different branches and even societies. In the manufacturing industry, value creation processes change as information and communication technologies merge with production processes. The change may enable efficiency gains and new business models. However, many firms still struggle to drive their digital transformation forward. To understand the barriers which hinder or even stop DT is essential for the successful transformation. Our study aims at identifying the barriers on the basis of 46 expert interviews. These practical insights are further used to develop a research agenda. To determine the research gaps, we conduct a literature review on the topics mentioned by the interviewees. Thus, we contribute by first of all identifying major barriers which can support firms by reflecting their DT. Moreover, we give an outlook for researchers on possible future exploration. So, we bring together perspectives from research and practice.

Digital Transformation (DT) is an essential trend for manufacturing companies as digitalization of the value chain affects the entire company. Strategic management functions should consider DT technologies and their impact on assets and resources. The knowledge of the possible factors that influence DT positively may ease a realization of gains due to DT. In 20 qualitative interviews, we examined critical factors for DT's success in manufacturing companies. Based on the IS success model of DeLone and McLean, we have derived success factors that can trigger DT success. The success factors describe the three major dimensions in which DT takes place: technology, organization, and environment. The results show that DT can only be successful if companies collaborate with customers, suppliers and also other firms from the branch. A cultural change is necessary to enable an agile working environment as well as more interdisciplinary activities. It becomes evident that the choice of technology is essential. However, driving only technology forward is not enough to gain benefits from DT.

Digital transformation (DT) describes technology-based improvements in business processes, business models, and customer experience. It promises efficiency gains for industrial enterprises. Nonprofit organizations also expect advantages from DT. However, barriers hinder realizing all its possible advantages in both sectors. If decision-makers recognize the potential barriers, they can reflect upon these challenges and take well-coordinated countermeasures. Orienting towards a Straussian grounded theory approach, a framework of barriers is developed with data of two diverse sectors: industry and nonprofit. According to the framework pre-conditions such as profit-orientation and size shape the possibilities to tackle different barriers. In general, the DT process in the industry-sector has been slowed down by barriers. Whereas, nonprofit organizations often take the view that they are not in a DT process at all. This might be due to limited individual and organizational perspectives. Especially, NPOs have to work on their recruitment of skilled volunteers to challenge this view.

The ongoing process of digital transformation in manufacturing – known as Industry 4.0 - hauls fundamental change. The whole value chain of enterprises is affected. As the digital transformation of businesses is still ongoing, many enterprises struggle with the challenges arising. This paper aims to show these struggles but also to contribute by analyzing how enterprises are transforming. We take a phenomenological view of the ongoing transformation. To get in-depth insights, we conducted and analyzed 18 interviews with 10 companies. For most companies, the digital transformation starts in operations with the vision of building a smart factory. Other primary and support activities also need to transform. These essential changes lead to restructuring and extensions of the strategy of manufacturing companies. Following these changes, companies will not need to choose either cost advantage or differentiation as a strategy but instead can do both.

The current knowledge of the effects of the physical environment on learners' behaviour in collaborative problem-solving tasks is underexplored. This paper aims to critically examine the potential of multimodal learning analytics, using new data sets, in studying how the shapes of shared tables affect the learners' behaviour when collaborating in terms of patterns of participation and indicators related to physical social interactions. The research presented in this paper investigates this question considering the potential interplay with contextual aspects (level of education) and learning design decisions (group size). Three dependent variables (distance between students, range of movement and level of participation) are tested using quantitative and qualitative analyses of data collected using a motion capture system and video recordings. Results show that the use of round tables (vs rectangular tables) leads to higher levels of on-task participation in the case of elementary school students. For university students, different table shapes seem to have a limited impact on their levels of participation in collaborative problem solving. The analysis shows significant differences regarding the relationship between group size and the distance between students, but there is no substantial evidence that group size affects the level of participation. The findings support previous research highlighting the importance of studying the role of the physical environment as an element of learning design and the potential of multimodal learning analytics in approaching these studies.

Intelligent telehealth system (ITS) provides patients and medical institutions with a lot of convenience, medical institutions can achieve medical services for patients in time through monitored health data. However, as the scope of people?s daily activities extends, the traditional single-server architecture is no longer applicable. To deal with this problem, a multi-server architecture has been proposed recently while there remains security and privacy challenges, including handover authentication. In this paper, we investigate a blockchain-assisted handover authentication and key agreement scheme for ITS in a multi-server edge computing environment. Specifically, we first propose a novel handover authentication model of ITS with multi-server edge computing architecture. Second, the proposed handover authentication scheme allows the authenticated server to assist users subsequently authenticate with other server, thereby achieving interactions with the server anytime and anywhere with low overhead. Finally, blockchain technology and strong anonymity mechanism are introduced to protect users? privacy strictly. To our best knowledge, the proposed scheme is the first in the literature to provide efficient authentication, strict anonymity and computational load transfer simultaneously. The security analysis and performance evaluation show that our scheme can not only satisfy the security requirements but also achieve higher efficiency in computation and communication cost.

This study extended the knowledge over the improvement of the control performance for a seven degrees-of-freedom (7DOF) humanoid arm. An improved adaptive Gaussian radius basic function neural network (RBFNN) approach was proposed to ensure the reliability and stability of the humanoid arm control. Considering model uncertainties, the established dynamic model for the humanoid arm was divided into a nominal model and an error model. The error model was approximated by the RBFNN learning to compensate the uncertainties. The contribution of this study mainly concentrates on employing fruit fly optimization algorithm (FOA) to optimize the basic width parameter of the RBFNN, which can enhance the capability of the error approximation speed. Additionally, the output weights of the neural network were adjusted using the Lyapunov stability theory to improve the robustness of the RBFN-based error model. The simulation and experiment results demonstrate that the proposed approach is able to optimize the system state with less tracking errors, regulate the uncertain nonlinear dynamic characteristics, and effectively reduce unexpected interferences.

Industrial Internet is applicable across a broad industrial spectrum including manufacturing, aviation, road and rail transport, power, oil and gas, healthcare, smart cities and buildings. Some of the major impacts of the Industrial Internet include the development of new and innovative services and products, which in turn also has economic benefits. The purpose of this special issue is to bring together research studies proposing novel techniques, algorithms, models, and solutions to address challenges such as interoperability, security, and privacy associated with Industrial Internet, blockchain and Cyber-physical systems.

The purpose of this paper is to show the technical feasibility of implementing our mobile context-aware learning schedule (mCALS) framework as a software application on a mobile device using current technologies, prior to its actual implementation. This process draws together a set of compatible mobile and context-aware technologies at present and can be used as a reference point for implementing generic mobile context-aware applications. Our mCALS framework retrieves the learner’s location and available time contexts via the built-in learning schedule (i.e. electronic organizer) on a mobile device. These contexts together with the learner’s learning styles and knowledge level (on a selected topic) are used as the basis for the software application to suggest learning materials that are appropriate for the learner, at the time of usage. This retrieval approach attempts to eliminate the use of context-aware technologies and the need to directly request the user to enter context information at the time of usage. We aim to develop a fully functional prototype of this framework for learners to plan their individual as well as social learning activities amongst one another in order to make their individual learning processes collaborative and as a way to enhance individual and social learning experiences.

In recent times, there has been an uptake of mobile learning (hereafter, abbreviated as m-learning), i.e. learning with mobile technologies, especially in the form of mobile apps. Apps are particularly useful as they are small applications usually with a single-purpose and allow the users to view learning content offline from their mobile devices. They can be a good supplement to formal learning/training and can be very efficient informal learning tools, allowing learners to learn anytime and anywhere. In this paper, we present our "Learning Java" app, which was designed based on the theoretical framework "context-aware personalized m-learning application with m-learning preferences" (Yau and Joy 2011). The app utilizes a personalized user profile consisting of location, noise and time of day, as well as the learner's knowledge level. Additionally, an understanding of different m-learning preferences by learners is represented in our app as their individual user profile, for example, a learner may concentrate the best in a quiet library and the app will select appropriate (more difficult and longer) material based on this information, as opposed to shorter and easier materials. Video materials are also used by learners. This app was tested by 40 volunteers; 10 of which completed a long questionnaire regarding the usage of the app in terms of personalized user profile, context-awareness factors and whether the app helped increase their motivation for learning and their learning effectiveness for the subject. The results highlighted that participants could optimize their spare times for most effective learning (e.g. video-watching with headphones) in busy and/or noisy environments. Findings also showed other chosen learning strategies by learners to make their learning more effective. Future work includes (1) extending the app for other subjects and disseminating it for use by remote learners, for example, those who are situated in developing countries without frequent access to wireless internet and/or educational materials, and (2) including learning analytical support to students to enhance their study success.

Advancements in technology, such as smartphones and wearabledevices, can be used for collecting movement data through embeddedsensors. This paper focuses on linking Parkinson’s Disease severitywith data collected from mobile phones in the mPower study. As referencefor symptoms’ severity, we use the answers provided to part 2 ofthe standard MDS-UPDRS scale. As input variables, we use the featurescomputed within mPower from the raw data collected during 4 phonebasedactivities: walking, rest, voice and finger tapping. The features arefiltered in order to remove unreliable datapoints and associated to referencevalues. After pre-processing, 5 Machine Learning algorithms areapplied for predictive analysis. We show that, notwithstanding the noisedue to the data being collected in an uncontrolled manner, the regressedsymptom levels are moderately to strongly correlated with the actualvalues (highest Pearson’s correlation = 0.6). However, the high differencebetween the values also implies that the regressed values can not beconsidered as a substitute for a conventional clinical assessment (lowestmean absolute error = 5.4).

We extended the Mobistudy app for clinical research in order to gather data about Parkinson’s motor and non-motor symptoms. We developed 5 tests that make use of the phone’s embedded sensors and 3 questionnaires. We show through data collected by healthy individuals simulating PD symptoms that the tests are able to identify the presence of symptoms.

In many situations, it is essential to analyze a nonstationary signal for sensing whose components not only overlapped in time-frequency domain (TFD) but also have different durations. In order to address this issue, an improved separation method based on the time-frequency distribution is proposed in this paper. This method computes the time-frequency representation (TFR) of the signal and extracts the instantaneous frequency (IF) of components by a two-dimensional peak search in a limited area in which normalized energy is greater than the set threshold value. If there is more than one peak from a TFR, IFs of components can be determined and linked by a method of minimum slope difference. After the IFs are obtained, the improved time-frequency filtering algorithm is used to reconstruct the component of the signal. We continue this until the residual energy in the TFD is smaller than a fraction of the initial TFD energy. Different from previous methods, the improved method can separate the signal whose components overlapped in TFR and have different time durations. Simulation results have shown the effectiveness of the proposed method.

Labelling user data is a central part of the design and evaluation of pervasive systems that aim to support the user through situation-aware reasoning. It is essential both in designing and training the system to recognise and reason about the situation, either through the definition of a suitable situation model in knowledge-driven applications, or through the preparation of training data for learning tasks in data-driven models. Hence, the quality of annotations can have a significant impact on the performance of the derived systems. Labelling is also vital for validating and quantifying the performance of applications. In particular, comparative evaluations require the production of benchmark datasets based on high-quality and consistent annotations. With pervasive systems relying increasingly on large datasets for designing and testing models of users' activities, the process of data labelling is becoming a major concern for the community. In this work we present a qualitative and quantitative analysis of the challenges associated with annotation of user data and possible strategies towards addressing these challenges. The analysis was based on the data gathered during the 1st International Workshop on Annotation of useR Data for UbiquitOUs Systems (ARDUOUS) and consisted of brainstorming as well as annotation and questionnaire data gathered during the talks, poster session, live annotation session, and discussion session.

Complexity and sophistication among multimedia-based tools have made it easy for perpetrators to conduct digital crimes such as counterfeiting, modification, and alteration without being detected. It may not be easy to verify the integrity of video content that, for example, has been manipulated digitally. To address this perennial investigative challenge, this paper proposes the integration of a forensically sound push button forensic modality (PBFM) model for the investigation of the MP4 video file format as a step towards automated video forensic investigation. An open-source multimedia forensic tool was developed based on the proposed PBFM model. A comprehensive evaluation of the efficiency of the tool against file alteration showed that the tool was capable of identifying falsified files, which satisfied the underlying assertion of the PBFM model. Furthermore, the outcome can be used as a complementary process for enhancing the evidence admissibility of MP4 video for forensic investigation.

In recent years, with the development of computation capability in devices, companies are eager to investigate and utilize suitable ML/DL methods to improve their service quality. However, with the traditional learning strategy, companies need to first build up a powerful data center to collect and analyze data from the edge and then perform centralized model training, which turns out to be inefficient. Federated Learning has been introduced to solve this challenge. Because of its characteristics such as model-only exchange and parallel training, the technique can not only preserve user data privacy but also accelerate model training speed. The method can easily handle real-time data generated from the edge without taking up a lot of valuable network transmission resources. In this paper, we introduce an approach to end-to-end on-device Machine Learning by utilizing Federated Learning. We validate our approach with an important industrial use case in the field of autonomous driving vehicles, the wheel steering angle prediction. Our results show that Federated Learning can significantly improve the quality of local edge models and also reach the same accuracy level as compared to the traditional centralized Machine Learning approach without its negative effects. Furthermore, Federated Learning can accelerate model training speed and reduce the communication overhead, which proves that this approach has great strength when deploying ML/DL components to various real-world embedded systems.

With the increasing attention on Machine Learning applications, more and more companies are involved in implementing AI components into their software products in order to improve the service quality. With the rapid growth of distributed edge devices, Federated Learning has been introduced as a distributed learning technique, which enables model training in a large decentralized network without exchanging collected edge data. The method can not only preserve sensitive user data privacy but also save a large amount of data transmission bandwidth and the budget cost of computation equipment. In this paper, we provide a state-of-the-art overview of the empirical results reported in the existing literature regarding Federated Learning. According to the problems they expressed and solved, we then categorize those deployments into different application domains, identify their challenges and then propose six open research questions. 

In recent years, with more edge devices being put into use, the amount of data that is created, transmitted and stored is increasing exponentially. Moreover, due to the development of machine learning algorithms, modern software-intensive systems are able to take advantage of the data to further improve their service quality. However, it is expensive and inefficient to transmit large amounts of data to a central location for the purpose of training and deploying machine learning models. Data transfer from edge devices across the globe to central locations may also raise privacy and concerns related to local data regulations. As a distributed learning approach, Federated Learning has been introduced to tackle those challenges. Since Federated Learning simply exchanges locally trained machine learning models rather than the entire data set throughout the training process, the method not only protects user data privacy but also improves model training efficiency. In this paper, we have investigated an advanced machine learning algorithm, Deep Neural Decision Forests (DNDF), which unites classification trees with the representation learning functionality from deep convolutional neural networks. In this paper, we propose a novel algorithm, AF-DNDF which extends DNDF with an asynchronous federated aggregation protocol. Based on the local quality of each classification tree, our architecture can select and combine the optimal groups of decision trees from multiple local devices. The introduction of the asynchronous protocol enables the algorithm to be deployed in the industrial context with heterogeneous hardware settings. Our AF-DNDF architecture is validated in an automotive industrial use case focusing on road objects recognition and demonstrated by an empirical experiment with two different data sets. The experimental results show that our AF-DNDF algorithm significantly reduces the communication overhead and accelerates model training speed without sacrificing model classification performance. The algorithm can reach the same classification accuracy as the commonly used centralized machine learning methods but also greatly improve local edge model quality.

Machine Learning (ML) and Artificial Intelligence (AI) have increasingly gained attention in research and industry. Federated Learning, as an approach to distributed learning, shows its potential with the increasing number of devices on the edge and the development of computing power. However, most of the current Federated Learning systems apply a single-server centralized architecture, which may cause several critical problems, such as the single-point of failure as well as scaling and performance problems. In this paper, we propose and compare four architecture alternatives for a Federated Learning system, i.e. centralized, hierarchical, regional and decentralized architectures. We conduct the study by using two well-known data sets and measuring several system performance metrics for all four alternatives. Our results suggest scenarios and use cases which are suitable for each alternative. In addition, we investigate the trade-off between communication latency, model evolution time and the model classification performance, which is crucial to applying the results into real-world industrial systems.

With the development and the increasing interests in ML/DL fields, companies are eager to apply Machine Learning/Deep Learning approaches to increase service quality and customer experience. Federated Learning was implemented as an effective model training method for distributing and accelerating time-consuming model training while protecting user data privacy. However, common Federated Learning approaches, on the other hand, use a synchronous protocol to conduct model aggregation, which is inflexible and unable to adapt to rapidly changing environments and heterogeneous hardware settings in real-world scenarios. In this paper, we present an approach to real-time end-to-end Federated Learning combined with a novel asynchronous model aggregation protocol. Our method is validated in an industrial use case in the automotive domain, focusing on steering wheel angle prediction for autonomous driving. Our findings show that asynchronous Federated Learning can significantly improve the prediction performance of local edge models while maintaining the same level of accuracy as centralized machine learning. Furthermore, by using a sliding training window, the approach can minimize communication overhead, accelerate model training speed and consume real-time streaming data, proving high efficiency when deploying ML/DL components to heterogeneous real-world embedded systems.