Abstract: Comtrade PMO (Project Management Office) established an expert group for PMO Education, having a vision about improving company efficiency by extending a set of PM’s (Project Manager) skills, applying new processes and standards while embracing new learning habits of PM professionals. Education of project managers in the software industry should take into consideration not only market changes and demands but also employees’ needs as well as internally recognized areas for improvements. It must be flexible to provide proper aligning with frequent market changes while affecting corporate efficiency in a measurable manner. Due to changed learners, habits, different content, and communication channels must be established to provide equally important self-paced learning and productive collaboration. This paper shows the whole process flow from initial roadmap to implementation, state of PMO Education, problem-solving framework as well as actionable KPIs (Key Performance Indicator), which unambiguously show significant achievements in the excellence of Comtrade PM Experts.

Abstract: This research addresses the anatomic prosthesis modeling problem based on contour reconstruction from Computed Tomography images. The new neural network techniques permit predicting the geometric attributes of bone failures that cannot be obtained by symmetry. This paper presents a method based on Generative Adversarial Network (GAN) applied to build anatomic 3D geometric models of a skull prosthesis. The method performs the calculation of geometric descriptors as parameters to fit a curve on each Computed Tomography slice. These features can be estimated by the network who will be trained with a database of descriptors. The curve descriptors are formed by the fixed points around the edge that better represent a geometry, and a curve section can be described using the Cubic Splines method in order to complete the missing information on an open edge. The curve that fits a missing region to each CT slice can be superimposed to define the stack of images to build a 3D virtual CAD model. A prototype software system was implemented in Python in order to calculate the edge parameters and generate the computational model to printing. The results evaluate the similarity measured between the estimated and real coordinates and demonstrate the feasibility of the proposed method.

Abstract: This paper is devoted to the problem of distributed multi-agent blind macro calibration of large Wireless Sensor Networks (WSN). The paper starts with establishing strong connections between Cyber-Physical Systems (CPS) and WSN. General considerations of the problem of WSN calibration are provided. The central point of the paper is presentation of an original algorithm for multi-agent recursive blind macro calibration based on consensus, including an exact proof of convergence. Simulation results are presented as an illustration of the properties of the algorithm.

Abstract: This paper presents an implementation of the Software Vulnerability Management System. The main goal of this system is to analyze data flow diagrams in order to discover potential security exploits. The security experts put great effort to discover vulnerabilities in different systems and to successfully prevent all possible attacks. They spend many working hours analyzing diagrams and targeting possible vulnerabilities because targeting them on time reduces the overall cost and time spent on the later phases of the software design. This system uses a knowledge base to find exploits on the different data flow diagrams and searches the National Vulnerability Database (NVD) to discover publicly known vulnerabilities that are associated with the diagram elements. This system brings a new type of element, that could be used on the diagram with the processes, external entities, and data stores. That element is called a complex process and it contains a new data flow diagram. This proposed system automates the analyzing process, simplifies threat modeling, and provides a readable report with the list of all found potential security exploits. It was designed as a website with a modest user interface that is easy to use, even by non-experts.

Abstract: Web applications are one of the most vulnerable systems nowadays. Everybody uses them in everyday life and most often they handle sensitive information from users, which makes them very tempting to malicious users for crashing them or steal their data. Testing web applications against various cybersecurity attacks can lower the risk of malicious user to perform some action that can harm the system itself or its users. Adequate test data can reveal these vulnerabilities, but traditional testing methods can not determine which test sets are appropriate. It is not uncommon that testing comes down to brute force testing. Also, these techniques do not take care of the correlation between different parameters which, under certain conditions, affect the reliability of the system. In this paper, we apply the idea of a combinatorial testing technique to generate adequate test data sets for testing web applications against known cybersecurity vulnerabilities. We used ACTS for automatic test set generation. The results show that these test sets are effective and can reveal security issues. Also, the number of necessary test cases is reduced, which directly affects the time and money needed to adequately test the software. Keywords software testing, combinatorial testing, web security.

Abstract: Nowadays, smart homes have the goal of providing an easier and more comfortable lifestyle. For that purpose, smart homes contain a vast number of electronic appliances and consume electricity from the electricity grid or renewable energy sources such as solar panels. Smart scheduling of these electrical appliances to lower the consumed energy while keeping the residents satisfied is a complex problem. One solution would be the implementation of an intelligent Home Energy Management System (HEMS) that automatically determines the best schedule. The focus of this paper is the implementation of a framework that could facilitate the training of a smart HEMS, using Siebog a Multi-Agent based system. To ensure resident satisfaction, we propose a system that enables residents to define their priorities in a narrative form, e.g. “The dishes should be washed till 7 AM”. Additionally, to provide smart and optimal energy consumption, as a part of the proposed system, we have implemented a cost function metric for machine-learning-based optimization that combines the two goals of resident satisfaction and lowering electricity consumption.

Abstract: Achieving energy efficiency is crucial in order to improve environment and life quality. Therefore, within the residential sector, research efforts on advanced approaches to steer end users to save the energy and adapt their consumption are widely spread. One of recent approaches being used is creating a competitive environment which would stimulate the users to improve their habits. The main idea of this paper is proposing a data-driven machine learning clustering approach for ranking users depending on their consumption measurements and other relevant data collected from a real-world pilot in France. Within this paper, exploited data, implementation details and results will be presented.

Abstract: With the aim of improving ecological interest, the share of renewable energy sources (RES) in the energy production is to be increased. Nonetheless, that growth adversely influences the grid’s instability, as a result of the dependency between the RES production and weather conditions. Therefore, with the aim of providing a stable energy system, it is necessary to plan the consumption in advance with respect to the availability of RES production. This paper is focused on comparing current SoA approaches for two different renewable energy sources, photovoltaic panels and solar thermal collector, using real world data from Denmark and Spain.

Abstract: With the IoT industry rapidly growing and evolving, smart home concepts powered by distributed IoT sensors are becoming more and more popular amongst residential users. This transformation has created an opportunity for novel energy management solutions with the main goal of improving energy efficiency. A key element in this process is being able to predict energy consumption with this paper aiming to extend the current state of the art by analyzing real measurements obtained from smart sensors through their distribution, statistical characteristics and suitability as predictors in a forecasting algorithm.

Abstract: It is a frequent case of cancellation or absence of patients on scheduled medical care. This behavior results in additional costs and unnecessarily long waiting times for patients arriving on scheduled appointments. This paper attempts to minimize the consequences of the problem by predicting patients' cancellations and absences by applying logistic regression, enabling employees at healthcare institutions to optimize their existing schedules and reduce the number of empty appointments. Prediction is made using a model that is trained on actual data collected in several health institutions in Serbia from 2010 to 2019.

Abstract: The amount of geospatial data generated globally, together with the necessity for increased interoperability of these data, call for innovative solutions for global geospatial reference frames. Discrete Global Grid Systems are a class of spatial reference systems that use hierarchical tessellation of cells to partition and address the globe without gaps or overlaps. The specific properties of DGGS make them important candidates for future standard geospatial reference frames and fuel further efforts to investigate their potential for organization, exchange and processing such data. In this paper, we focus on DGGSs based on the so-called spherical cube mapping technique and present some first results of how these can be optimized to serve as global reference frames for large volumes of gridded geospatial data.

Abstract: Sport and rehabilitation are two very important research areas where biomechanical systems and applications are used. The use of sensors and smart equipment enables the measurement of motion parameters that were previously unknown or not available. Sensors can produce large amounts of data that need to be stored for later processing and analysis. It is also important that the cloud platform architecture is structured yet flexible enough to allow different types of experiments and applications to use a single cloud platform to some extent. This paper presents a universal solution that addresses these issues. We present the architecture for the proposed cloud platform that can accommodate all types of sensors, support different applications and measurements, track users through different tests, and produce standardized output types and reports.

Abstract: Irrational energy consumption and waste of energy are two of the main problems that we face today. A lot of the energy is wasted by people in their homes or offices because they are not paying much attention to their energy consumption. In order to increase energy efficiency of the general population and to raise general awareness, a system based on Internet of Things (IoT) is developed to influence users to be more energy efficient. Because of the great number of different entities in the buildings and their complex relationships, a semantic ontology is used to store contextual knowledge about spatial arrangements of the rooms, devices and sensors in the buildings.

Abstract: Software Product Line (SPL) engineering is a paradigm which allows reducing development time, effort, and costs for development of products with the same core features and some variations needed for every client that purchases the product. Instead of writing the variations code from scratch, we can follow the Model-Driven development approach, which aims to generate software from design models automatically. Incorporating Model-Driven Software development in an existing large scale software product line (SPL) can be challenging and full of obstacles due to constant development and changes in the SPL core and product architecture. The introduction of the Model-Driven approach to such solutions often must be done in an iterative and incremental manner to embrace the changes. In order to achieve this goal, it is necessary to fulfill two preconditions: the existence of the domain-specific modeling language and transformation programs for automatic code generation from a model. This paper presents a meta-model and a code generator that enables rapid development and customization of the SPL applications. Our solution enables the core product line to be automatically expanded in any segment (e.g. method, data structure, etc.).

Abstract: We investigate the computational complexity of tracking 3D orientation (3DO) using gyroscope angular velocity measurements around its three sensitivity axes. These three rotations, obtained at every measurement step, are simultaneous and, due to rotation non-commutativity, interpreting them as sequential, i.e., Euler rotations, yields a systematic error in the estimated 3DO. As this error grows with the angle of rotation, an efficient approach for its reduction is to shorten the measurement interval, demanding a larger number of measurement and hence computational steps. We show that from the computational complexity point of view, for similar levels of result accuracy, the simultaneous interpretation of gyroscope measurements is superior to the sequential. Experimental results obtained using a dedicated microcontroller and relying on a specially developed, computationally optimized implementation, show that for the largest rotation angle considered, i.e., 3.67°, execution time for the simultaneous interpretation is 12 times shorter than for the sequential. Aiming to achieve computational efficiency and relevant comparison of both interpretations, rotation matrices were used when calculating 3DO after each measurement step and the rotation quaternions were used when combining multiple consecutive measurements in a single rotation composite.

Abstract: Service Oriented Architecture (SOA) is the most popular and widely adopted architectural style for implementing and integrating large-scale heterogeneous systems. While it has a lot of forms, microservices were established as the preferred way of building distributed SOA applications. Such architectures tend to have a lot of asynchronous REST communication. Their distributed nature, however, often results in complex design decisions that address the issue of reliable messaging. In this work, we present a self-adaptive agent that guarantees reliable RESTful messaging. The agent can be integrated into existing services, thus reduce the overall complexity of the entire infrastructure.

Abstract: Precise tracking of the tennis player’s location gives important information for the tennis game analysis. We designed a prototype system that uses an ultra-wideband (UWB) positioning technology for players tracking on the tennis court and wireless inertial sensors for detecting the ball impacts. We verified the accuracy of the UWB positioning technology in the tennis court by the high-precision optical tracking system. The results of the field tests confirmed that UWB technology is useful for player tracking in tennis coaching applications.

Abstract: Measurement of the relevant basic and sport specific physical abilities of athletes is one of the most important elements in the system of sport. Agility, i.e. the ability to rapidly change direction is one of these abilities and various agility tests are regularly included in sport test batteries. This paper presents a state-of-the-art approach to measuring agility using infrared and kinematic sensors. The combination of precise timing, position detection and motion recording enables the acquisition of previously inaccessible kinematic and temporal variables. The evaluation of the athletes' agility can be more objective and acquired with much greater certainty than it would be possible with current equipment and methods. The presented sensor system can be used in different sport settings and types of tests, including the frequently used T-test, which serves as our case study. Keywordsagility, photocell, IMU, sensor system, synchronization, sensor fusion, T-test

Abstract: Branch predictors are one of the most significant components in the contemporary pipelined processors. They are used to predict which instruction should be executed right after a branch instruction. Prediction itself must be made before completion of the branch instruction. This requirement makes an accuracy the main characteristic of a branch predictor. Accuracy of a branch predictor naturally must be very high to avoid wasted processor cycles while executing instructions which are not part of the correct execution path. Additionally, in case of a misprediction processor must engage recovery procedure in order to return back to the correct execution path which reduces performance of the processor. Branch predictors use various algorithms and structures to calculate a prediction consuming certain time in the process. This time is called a branch predictor latency. There are two types of branch predictor latency. First type is time needed for prediction calculation itself and second type is time needed to update internal state of branch predictors. This paper focuses on time needed for prediction calculation and ways this type of branch predictor latency could be mitigated. There is a proposition of a mechanism that initiates prediction calculation in advance i.e. before branch instruction has even been fetched for execution by the processor. Proposed mechanism consists of two different branch predictors and internal structures for information buffering. First branch predictor is slow in terms of time needed for prediction calculation and has very high accuracy whilst second branch predictor is fast but less accurate. Depending on the execution traces used for testing proposed mechanism achieves accuracy from 76 to 99 percent. Index terms: branch predictor, prediction calculation latency, pipelined processor

Abstract: The role of cryptocurrencies in the alternative world economy has been steadily increasing since they emerged in 2009. However, understanding their actual influence and performance can be a difficult task. In this paper, we present an architecture for real-time cryptocurrency data processing and analysis based on the Lambda architectural approach. The proposed architecture offers both batch and stream data processing of cryptocurrency transactions and blocks, as well as analyzing various sorts of trends in a blockchain network and an exchange market. The proposed architecture is modular and thus easies the implementation of loosely coupled components and also provides several benefits for cryptocurrency analysis: real-time monitoring of blockchain events and mining statistics, cryptocurrency buying and selling trends, as well as social media events related to cryptocurrency reputation.