Abstract: The reliability of IoT networks gained its significance during the past several years, mainly due to the harmful coexistence with other technologies operating in the same 2.4 GHz frequency band. IoT applications, using ZigBee, have much lower power than WiFi, the most widespread technology in this frequency band. Such a coexistence results in a significant packet loss in ZigBee networks, which makes them unreliable. The sensor data driven design directions, developed using the massive online IoT testbed SmartSantander, are given in this paper. The interference problems are identified and described, which is the most important step towards the future performance improvement of IoT applications.

Abstract: Supporting mining processes by modern information and communications technologies is emerging area. We have previously enabled monitoring of some important parameters in open pit mining and set an integral framework for sustainable mining. In this paper, we extend our activities in monitoring phase according to the proposed framework and analyze various ways of data collecting. We show that combining data from proprietary, open data and public sources, sufficient set of data for sustainable control of open pit mining process can be established.

Abstract: In this paper, we will analyze benefits of two popular private blockchain platforms, Ethereum and Hyperledger Fabric. Blockchain is decentralized transaction and data management technology and is said that its impact will be similar to the impact Internet had. Blockchain can be implemented in many industries, but the question is: is it needed, regarding scalability, business value and performance? The goal of this paper is to present analysis results regarding business value of both blockchain platform.

Abstract: With the expressive amount of terms and concepts generated over the years for all the existing domains, a formalization and possible validation of such domains were given as necessity. In the medical field, for example, more precisely in the field of implantology, terms are used daily in different ways around the world. The standardization of such terms is extremely important for human and machine understanding, thus creating a semantic interoperability between both systems. Thus, the use of ontologies as a method of knowledge formalization has been an approach to benefit the human understanding for any real domain. Applying this method to the area of implantology, ontological basis comes to benefit the professional dental surgeon at the time of decision making in choosing which type of dental implant to be applied to each particular case. This method consists of determining which implant is to be applied before a numerous physiological parameters of the patient compared to structural parameters of existing dental implants. For this method to be a real application, the formalized ontologies must contain and supply all essential data related to this domain. Therefore, this project focuses on the formalization of an ontological basis aimed at the application of a dental implant before physiological human and implant variables by itself. In order to obtain the results, it is essential to keep in mind the real meaning of the semantics and interoperability concepts, which should be applied together to the formalization of an ontology. In addition, it is necessary to analyze the application of these concepts in the dental field so that a case study based on the formalized ontology can be applied later. The formalization of the ontology was made based on Protégé software created by specialists of the university of Stanford in California. The results showed that not only an efficient but also effective solution for such an application was obtained, making the decision of dental professionals simpler with lower failure rates and better acceptance of the patient's body in relation to osseointegration.

Abstract: The aim of this paper is to present an approach to quantitative analysis of syllable distribution in Serbian texts. The approach is based on creating a software that allows automatic processing of texts in Serbian, and which will have as a result the text with words divided in syllables, according to the syllabification rules for Serbian. Also, the program should give a detailed quantitative analysis of the processed text, which included the analysis of syllables by type, length, frequency, words, and text. In this paper some of the initial results are presented.

Abstract: Although the financial benefits are crucial for today's market oriented industry, environment protection is recognized as an ultimate goal for modern industry development. This goal is usually performed in the frame of energy efficiency. It has been shown that non-intrusive load monitoring (NILM) is not only beneficial for planning and optimization of production but also for consumption of energy. Overview of basic problems related to NILM, key idea, theoretical frames for realization of this task and an implemented algorithm are presented in this paper.

Abstract: Although the financial benefits are crucial for today's market oriented industry, environment protection is recognized as an ultimate goal for modern industry development. This goal is usually performed in the frame of energy efficiency. It has been shown that non-intrusive load monitoring (NILM) is not only beneficial for planning and optimization of production but also for consumption of energy. Overview of basic problems related to NILM, key idea, theoretical frames for realization of this task and an implemented algorithm are presented in this paper.

Abstract: Tourism is one of the highly dynamic areas which already extensively are using the available Internet technologies. The Semantic Web as the next generation web, gives opportunity of having background knowledge about the meaning of web sources stored in a machine-readable way using appropriate ontologies. Several publicly available tourism ontologies already exist today. However, information management solutions for tourism are still at an early stage from a semantic point of view. Concept of tourism “ecosystems” gives opportunity to implement successful tourism sector strategy and address the system-level challenges that lead to build up the institutions that support tourism and solving the problems which includes inadequate physical and social infrastructure and insufficient attention to environmental issues. Based on previous principles authentic tourism ecosystem ontology is created and presented in this paper. Open source software OWLGrEd was used for ontology creation and a set of mutually compatible software tools for creating and implementation of semantic solutions, such as WAMP server, ARC2, PHP, MySQL, and SPARQL. Proposed conceptual solution can be used for development of the next generation tourism information systems.

Abstract: IoT system gathers data from physical devices integrated into a data exchange network. Upon analysis, intelligent systems or humans make decisions based on the analysis performed. The quality of decision made depends on the trustworthiness of the data sources. One of the key elements in the decision-making process is trustworthy interaction between IoT entity and time information. In order to make timely decision, it is necessary to be certain about the time of the data gathering. The authors apply PKI technology in order to secure trust between all parties in the communications and timestamping technology to verify digital data in specific time in trustworthy and verifiable way.

Abstract: Software qGrains is intended for estimation of hydraulic conductivity based on granulometry analysis of samples taken from exploration drills. Using qGrains, it is possible to make space classification, sistematization and schematization of sediments based on choosen percentage involvement or choosen grain-size. This is a usefull part of preparation for a higher level of hydrogeological interpretation, for example making hydrodinamical model.

Abstract: This paper explains the production of a personalized child orthosis with 3D printing technology. In agreement with the doctor’s specialists, an appropriate candidate was selected. The components of a child's orthosis have been modelled and adapted based on the scans from candidate acquired with 3D scanning technology. The 3D scanning process with OrtoSCAN digitizer and component modelling in Autodesk Fusion 360 are explained. Customized child orthosis components were made using two 3D printers: Prusa i3 P3Steel and Geeetech Prusa i3 Pro. Patient visit was followed by assembling of the orthosis. After final design and orthosis trial, the conclusion is that less modification on the shoulder part is needed and there is still room for further optimization of the orthosis. By using a generative design, it is possible to obtain an even more lighter customized orthosis that will satisfy the required mechanical properties and fulfill its function.

Abstract: The rapid process of urbanization of the population in the world, as well as increased application of mobile and ubiquitous computing in all spheres of human activity, caused the modernization of the conditions of transport and traffic. Smart devices with the ability to collect data from the context of vehicles and passengers provide the possibility of processing them in order to increase the quality of transport services. If this information is publicly available and used to form collective intelligence in order to improve the transport conditions for residents of the cities, then they contribute to the areas of smart traffic in the concept of smart cities. The paper describes a smart transport service, which provides the possibility of monitoring the vibrations in the vehicle, in order to improve the level of comfort when driving in urban transport. The client-server application collects information about vibrations, and the result of their processing is publicly available estimations of routes, according to the criterion of driving comfort.

Abstract: Cluster orchestration systems achieve high resource utilization, efficient task-packing, process-level isolation. They support high-availability applications with features that minimize fault and recovery time. These systems helped large internet companies like Google, Amazon, and Facebook to satisfy all their applications and big data workloads. With this ability, these systems can run a dozen of tasks at scale, to satisfy all operational needs. But current systems are not that easily applicable in domains with constant data flow with processing on the very edge of the network. We present a system for edge computing, with efficient resource utilization, orchestration, and taskpacking. With these properties, a system provides the ability to run user-defined tasks in clusters defined on the edge of the network for constant data acquisition, manipulation, and processing

Abstract: Road design is a complex task and despite a rich set of skills and road regulations, a number of design questions remain unanswered until after the road has been constructed. Our aim is to offer civil engineers an evaluation of a new road segment on a driving simulator early in the design phase. In this paper we determine the practical workflow for exporting the road design produced by civil engineers to a driving simulator. While industrial grade software focuses on a feasible, cost-effective road design from the construction point of view, a simulator requires, among others, an optimized 3D model for performance, compelling visual details and semantic road information for autonomous traffic. In the workflow we explicitly use open source or freely available, cross platform, and scriptable tools and formats. Finally, we perform a showcase where test subjects evaluated the new road segment using the state of the art 4 DOF hardware simulator, an eye tracker, and biometric sensors to determine the driver’s stress level and response time.

Abstract: Agent-based systems provide users with multitude of software entities – agents, which can perform various tasks, while demonstrating autonomy, reactivity social ability and pro-activity. One of the key features of the agent technology is the agent mobility. This feature provides agents with the capability of moving from one computer to another, from one platform to another. Agent mobility is a basic part of the Siebog Agent Middleware since its beginnings. At first, agents in Siebog have been written in Java, and were able to migrate only to Java-based Siebog middleware. With the advent of JavaScript support in browsers, it was possible to have multitude of software agents running in browsers. Siebog added this feature with the introduction of the Radigost – a JavaScript support for both client-side (browsers) and server-side. However, the Radigost did not provide a full agent mobility. For example, it was not possible to migrate the server-written agent to the browser. This paper describes an implementation of a full agent mobility. This enhanced mobility now enables agents to freely move from browser to server, from server to another server, from server to browser, and from browser to browser.

Abstract: Phone calling is a common task during driving. With the purpose of increasing awareness about its negative effects among drivers in Slovenia, the Slovenian Traffic Safety Agency in collaboration with University of Ljubljana and Nervtech, performed a user study which explored the effects of phoning during driving on the driving performance and the driver. The study was performed with participants aged from 22 to 67 years old with different driving experiences. To ensure driving safety, the study was performed in a high-fidelity motion-based driving simulator, which enabled a realistic and comparable driving experience for all participants. It was additionally equipped with sensors for observing cardiovascular, electrodermal and pupil activity. The detection response task (DRT) was also used for the assessment of the attentional effects of increased cognitive load from performing a phone call. The driving performance was observed through changes in speed, acceleration, lane deviation, reaction times and safety distance. The driver’s state was observed through data on changes in galvanic skin response (GSR), skin temperature, response times, and pupil activity. The results for driving performance did not show any statistically significant changes in mean speed, however significant changes were found in mean acceleration and response times for trials with a phone call compared to trials with only driving. Also, the driver’s mean galvanic skin response (skin conductance) increased and DRT response times and hit rates decreased statistically significantly.

Abstract: This paper presents a tool easesthe process of configuring and developing HyperLedger Fabric Blockchain networks. The tool is intended tohelp the developers reduce the time needed for setup and focus more on the business logic implementation. The solution would providea domain specific language (DSL) which would lean on the code generator that produces configuration files for the Fabric network, as well as functions for interacting with the chaincode (smart contract). In the paper we present abstract syntax for the DSL as well as concrete textual syntax. All the provided concepts are specific for the Hyperledger Fabric network. Plans for future work include developing concrete graphical syntax and defining bigger set of concepts for modeling Fabric network

Abstract: This paper addresses the problem of generating images, called adversarial images, that can "fool" a neural network (NN) trained for image classification. The NN is considered to be fooled if it assigns the generated image with high confidence to one of its classes. In this paper a simple search based solution is presented, with emphasis on the fact that detailed knowledge about the NN’s internal structure and the type of data it was trained on is not needed. This highlights the security vulnerabilities NNs have in regards to the simplicity of algorithms that can fool them. In addition to fooling neural networks, the developed solution can be used for dataset augmentation with the goal of achieving better generalizations of the NN model. Also, the solution may be used for understanding complex NN models. The solution was verified by using NNs of different architectures and trained for different classification tasks, and it performed well in all cases. KeywordsArtificial neural network, adversarial data, convolutional neural network, search algorithm, image classification, machine learning

Abstract: In the article we present physiological signals that can be measured on drivers with the purpose of determining drivers’ state and behavior. The results of the analysis can be used in various fields of research including automated driving, driving simulations and driver profiling. Physiological signals of drivers can be divided into ten groups corresponding to ten major organ systems of the human body. Most important signals include ECG (electrocardiograph) and PPG (photoplethysmograph) from cardiovascular system, EGG (electrogastrograph) from digestive system, and EEG (electroencephalograph), GSR (galvanic skin response) and pupil diameter from nervous system. GSR and eye tracking have recently gained in popularity. EGG signal analysis also seems promising.

Abstract: SCORM 2004 as a predominant e-learning standard does not satisfy educational and training needs anymore. Towards the intention that the new one (xAPI) is going to be accepted by IEEE as a commonly accepted standard, the research is performed to evaluate the pros and contras for replacing the actual with the new one. The research has a focus on experiments made to acquire the influence of xAPI on different eLearning processes such as learning, serious gaming, testing, reporting about results and measuring of real competences.