Intelligent Distributed Computing XI

We consider the deployment of island-based evolutionary algorithms (EAs) on unstable networks whose nodes exhibit correlated failures. We use the sandpile model in order to induce such complex, correlated failures in the system. A performance analysis is conducted, comparing the results obtained in both correlated and non-correlated scenarios for increasingly large volatility rates. It is observed that simple island-based EAs have a significant performance degradation in the correlated scenario with respect to its uncorrelated counterpart. However, the use of self-\(\star \) properties (self-scaling and self-sampling in this case) allows the EA to increase its resilience in this harder scenario, leading to a much more gentle degradation profile.

An evolutionary algorithm (EA) is said to be spatially structured when its individuals are arranged in an incomplete graph and interact only with their neighbors. Previous studies argue that spatially structured EAs are less likely to converge prematurely to local optima. Furthermore, they have been initially designed for distributed computing and it is often claimed that their parallelization is simpler than the equivalent non-structured algorithm. However, most of the empirical studies on spatially structured EAs use a predefined and fixed population size, whereas the full potential of this or any other any kind of EA can only be explored if the population size is properly set. This paper investigates optimal population sizes of spatially structured EAs (cellular EAs, in particular) and the relationship between that size, convergence speed and the degree of the structuring network. EAs structured by regular graphs with different degrees have been tested on different types of fitness landscapes. We conclude that in most cases graphs with low degree require smaller populations to converge consistently to global optima. However, if the population size is properly set, EAs structured by graphs with higher degrees not only converge to global optima with high probability, but also converge faster.

In this work, a job-flow scheduling approach for Grid virtual organizations (VOs) is proposed and studied. Users and resource providers preferences, VOs internal policies, resources geographical distribution along with local private utilization impose specific requirements for efficient scheduling according to different, usually contradictive, criteria. With increasing resources utilization level the available resources set and corresponding decision space are reduced. In order to improve overall scheduling efficiency, we propose an anticipation scheduling heuristic. It includes a target (anticipated) pattern solution definition and a special replication procedure for efficient and feasible resources allocation. A proposed anticipation algorithm is compared against conservative backfilling variations using such criteria as average jobs response time (start and finish times) as well as users and VO economic criteria (execution time and cost).

The software defined networking has opened new opportunities for offering network resources to end users “as a service”. For these purposes a number of technologies have been proposed and implemented to enable easy definition and management of network resources dynamically. In this paper we provide an overview of software defined network and technologies used for identifying network topology. We present three approaches based on Dijkstra’s Shortest Path Algorithm and evaluate their performance in an experimental study.

The paper introduces ActLog, a rule-based language capable of specifying actions paraconsistently. ActLog is an extension of 4QL\(^{\!\text{ Bel }}\), a rule-based language for reasoning with paraconsistent and paracomplete belief bases and belief structures. Actions considered in the paper act on belief bases rather than states represented as sets of ground literals. Each belief base stores multiple world representations which can be though of as a representation of possible states. In this context ActLog’s action may be then seen as a method of transforming one belief base into another. In contrast to other approaches, ActLog permits to execute actions even if the underlying belief base state is partial or inconsistent. Finally, the framework introduced in this paper is tractable.

We consider a supply chain management problem where a business alliance of small capacity retailers needs to collectively select a unique supplier considering the assignment’s efficiency at both the alliance and retailers’ level. We model the alliance as a multi-agent system. For this model, we present a modified Vickrey auction algorithm with regret minimization and compare it experimentally with aggregation of preferences by voting and standard Vickrey auction. Through simulation, we show that the proposed method on average reaches globally efficient and individually acceptable solutions. The solutions are evaluated in terms of different social welfare values.

Self-localization is a basic service for Intelligent Transportation Systems (ITS) such as traffic regulation services. Most of the used techniques are based on integration of Inertial Navigation System (INS) and Global Positioning System (GPS). However, navigation through areas such as tunnels, where GPS coverage is vulnerable, obliges the use of a different approach. Based on this observation, we designed and implemented a lightweight cooperative positioning algorithm based on Adaptive Localization Protocol (ALP). In this paper, we apply our method as support to an intersection service for traffic regulation, in which a group of concurrent cars shares an intersection/critical section. We found that our algorithm improves car position and regulates the traffic.

Today, two major wireless technologies exist, LTE and WLAN. Greatest challenge is to decide which one to use when both are available. Process of moving a device from LTE to WLAN network is called vertical handover. Many factors could impact the handover decision, and a lot of algorithms exist with their advantages and disadvantages. Here, a new algorithm is proposed which is based on fuzzy logic and a new combination of input factors. Also, a large number of vertical handover could produce ping-pong effect and reduce user performance. According to this, a new solution for reducing ping-pong effect is proposed. A solution is based on graph theory and finding bridges in a graph.

The lack of a suitable toolchain for programming the interaction space with coordination languages hinders their adoption in the industry, and limits their application as core calculus, proof-of-concept frameworks, or rapid prototyping/simulation environments. In this paper we present the \(\texttt {ReSpecT}\mathbb {X}\) language and toolchain as a first step toward closing the gap, by equipping a core coordination language (ReSpecT) with tools and features commonly found in mainstream programming languages, improving likelihood of adoption in real-world scenarios.

Remote Piloted Aircraft Systems (RPAS) are operating in highly critical contexts and carry out a wide collection of complex mission tasks through the use of sensors. In this paper, we present a new agent-based architecture that handles sensors of these platforms. Today, the requirements of the platform in terms of autonomy, modularity, robustness and reactivity as well as the industrial constraints call for the design of a new multifunction system architecture. Such a design may rely on multi-agent paradigm since it is modular by design and the agents naturally bring autonomy and pro-activity to the system. This paper presents new and original contributions: (1) an original agentification of the system in the form of a multi-agent architecture that helps to capture the dynamic of the environment; (2) firsts results of the architecture’s simulation for autonomy and scheduling evaluation.

Regular building inspections are a key means of identifying defects before getting worse or causing a building failure. As a tool for building condition inspections, Unmanned Aerial Vehicles (UAVs) or drones offer considerable potential allowing especially high-rise buildings to be visually assessed with economic and risk-related benefits. One of the critical problems encountered in automating the system is that the whole process involves a very complicated and significant amount of computational tasks, such as UAV control, localisation, image acquisition and abnormality analysis using machine learning techniques. Distributed software agents interact and collaborate each other in complicated systems and improve the reliability, availability and scalability. This research introduces a ubiquitous concept of software-agents to a drone-based building inspection system that is applied to crack-detection on concrete surfaces. The architecture and new features of the proposed system will be discussed.

Environmental monitoring has been regarded as one of effective solutions to protect our living places from potential risks. Traditional methods rely on periodically recording assessments of observed objects, which results in large amount of hybrid data sets. Additionally public opinions regarding certain topics can be extracted from social media and used as another source of descriptive data. In this work, we investigate how to connect and process the public opinions from social media with hybrid observation records. Particularly, we study Twitter posts from designated region with respect to specific topics, such as marine environmental activities. Sentiment analysis on tweets is performed to reflect public opinions on the environmental topics. Additionally two hybrid data sets have been considered. To process these data we use Hadoop cluster and utilize NoSql and relational databases to store data distributed across nodes in share nothing architecture. We compare the public sentiments in social media with scientific observations in real time and show that the “citizen science” enhanced with real time analytics can provide avenue to nominatively monitor natural environments. The approach presented in this paper provides an innovative method to monitor environment with the power of social media analysis and distributed computing.

In today’s society, there is a compelling need for innovative approaches for the solution of many pressing problems, such as understanding the fluctuations in the performance of an individual when involved in complex and high-stake tasks. In these cases, individuals are under an increasing demand for performance, driving them to be under constant pressure, and consequently to present variations in their levels of stress. Human stress can be viewed as an agent, circumstance, situation, or variable that disturbs the normal functioning of an individual, that when not managed can bring mental problems, such as chronic stress or depression. In this paper, we propose a different approach for this problem. The EUStress application is a non-intrusive and non-invasive performance monitoring environment based on behavioural biometrics and real time analysis, used to quantify the level of stress of individuals during online exams.

Social networks are intensively and extensively used to exchange news and contents in real time. The lack of a global authority for assessing posts truthfulness however allows malicious to exhibit unfair behaviours; identifying methodologies to detect hoaxes and defamatory content automatically is therefore more and more required. Social networks as Facebook and Twitter provided specific solutions and general approaches were also developed; in this paper we present a general model that takes into account both post as well as users’ credibility, using a duplex network of acquaintances and credibility among users. First experiments show that it is possible to distinguish individuals who post non-truthful content through a combined analysis of both the news content and the reposts they get from their contacts.

The paper investigates different security data sources and analyzes the possibility of their sharing in a uniform data storage on the basis of the ontological approach. An ontological model of the uniform hybrid storage is suggested. A common technique for security data inference based on this approach is developed. The results of experiments with the suggested ontology to construct the security data storage are discussed.

Nowadays the energy generation strategy of almost every nation around the world relies on a strong contribution from renewable energy sources. In certain countries the relevance taken by wind energy is particularly high within its national production share, mainly due to its large-scale wind flow patterns. This noted potentiality of wind energy has so far attracted public and private funds to support the development of advanced wind energy technologies. However, the proliferation of wind farms makes it challenging to achieve a proper electricity balance of the grid, a problem that becomes further involved due to the fluctuations of wind generation that occur at different time scales. Therefore, acquiring a predictive insight on the variability of this renewable energy source becomes essential in order to optimally inject the produced wind energy into the electricity grid. To this end the present work elaborates on a hybrid predictive model for wind power production forecasting based on meteorological data collected at different locations over the area where a wind farm is located. The proposed method hybridizes Extreme Learning Machines with a feature selection wrapper that models the discovery of the optimum subset of predictors as a metric-based search for the optimum path through a solution graph efficiently tackled via Ant Colony Optimization. Results obtained by our approach for two real wind farms in Zamora and Galicia (Spain) are presented and discussed, from which we conclude that the proposed hybrid model is able to efficiently reduce the number of input features and enhance the overall model performance.

In this paper the use of convolutional neural networks (CNN) is discussed in order to solve four class motor imagery classification problem. Analysis of viable CNN architectures and their influence on the obtained accuracy for the given task is argued. Furthermore, selection of optimal feature map image dimension, filter sizes and other CNN parameters used for network training is investigated. Methods for generating 2D feature maps from 1D feature vectors are presented for commonly used feature types. Initial results show that CNN can achieve high 68% classification accuracy for the four class motor imagery problem with less complex feature extraction techniques. It is shown that optimal accuracy highly depends on feature map dimensions, filter sizes, epoch count and other tunable factors, therefore various fine-tuning techniques must be employed. Experiments show that simple FFT energy map generation techniques are enough to reach the state-of-the-art classification accuracy for common CNN feature map sizes. This work also confirms that CNNs are able to learn a descriptive set of information needed for optimal electroencephalogram (EEG) signal classification.

Complexity of IoT systems and tasks that are put before them require shifts in the way resources and service provisioning are managed. The concept of fog computing is introduced so as to enhance IoT systems scalability, reactivity, efficiency and privacy. In this paper we present fog computing solution with context aware decision-making procedures distributed between IoT cloud platform and IoT gateways. The solution performs decision-making for smart actuation, based on analysis of sensory data streams, and context informed fog computing resource and service provisioning management based on topology changes. The state-of-the-art mainly focuses either on smart actuation enabled through insightful data analysis and machine learning, or on managing fog system itself in order to improve performance and efficiency. Our solution showcases how one software framework can be used to achieve both. Proof of concept experiments executed on a fog computing testbed validate our solutions performance in improving resilience and responsiveness of the fog computing system in context of topology changes.

Numerous studies have linked lighting conditions to how well humans have performed their daily activities. We have designed a system for the purpose of increasing productivity by tailoring the ambient lighting for various tasks, but which at the same time can gain the trust of its users by exposing the way it “thinks” through computational argumentation. With the recent emergence of smart bulbs, we now have the means for creating a software system that is able to understand which task is performed and adapt to it accordingly. While the key role of this system is to make activities more pleasurable and easier to perform, it can also predict future activities and make suitable recommendations.

Cloud e-Learning (CeL) is a new paradigm for e-Learning, aiming towards using any possible learning object from the cloud in a smart way and generate a personalised learning path for individual learners. An issue that appears before the generation of the learning path through automated planning, is to filter a pool of resources that are relevant to the learners profile and desires in order to enhance their knowledge and skills at a higher cognitive level. In this paper, we present a Recommender System for Cloud e-Leaning (CeLRS) that uses hierarchical clustering to select the most appropriate resources and utilise a vector space model to rank these resources in order of relevance for any individual learner. We discuss the issues raised and we demonstrate how CeLRS works.

Anomaly Detection (AD) in distributed cloud systems is the process of identifying unexpected (i.e. anomalous) behaviour. Many approaches from machine learning to statistical methods exist to detect anomalous data instances. However, no generic solutions exist for identifying appropriate metrics for monitoring and choosing adequate detection approaches. In this paper, we present the CRI-Model (Change, Rupture, Impact), which is a taxonomy based on a study of anomaly types in the literatureand an analysis of system outages in major cloud and web-portal companies. The taxonomy can be used as an anlaysis-tool on identified anomalies to discover gaps in the AD state of a system or determine components most often affected by a particular anomaly type. While the dimensions of the taxonomy are fixed, the categories can be adapted to different domains. We show the applicability of the taxonomy to distributed cloud systems using a large dataset of anomaly reports from a software company. The adaptability is further shown for the production automation domain, as a first attempt to generalize the taxonomy to other distributed systems.

Subjection of wireless Internet of Things (IoT) devices to energy resource exhaustion attacks gets increasing importance. Being stealthy enough for an attack target and systems of its monitoring such attacks are capable to exhaust energy of the device in a relatively short period and thereby impair the function and availability of the device. The paper analyzes possible types of ERE attacks, proposes an intruder model regarding this kind of attacks and provides experimental studies on the basis of a developed use case.

Service discovery is a well-known but important aspect of dynamic service-based systems, which is rather unsolved for megascale systems with a huge number of dynamically appearing and vanishing service providers. In this paper we deduce requirements for service discovery in megascale systems and analyze existing approaches with these in mind. Shortcomings of existing solutions are explained and a novel solution architecture is presented. It is based on the idea that service description data can be subdivided into static and dynamic properties. The first group remains constant over time while the second is valid only for shorter durations and has to be updated. Expressive service queries rely on both, e.g. service location as example for the first and response time for the latter category. In order to deal with this problem, our main idea is to also subdivide the architecture into two interconnected processing levels that work independently on static and dynamic query parts. Both processing levels consist of interconnected peers allowing to auto-scale the registry dynamically according to the current workload. Finally, some parts of the ongoing implementation based on Jadex agent technology will be explained.

Since computing devices in IoT tend to have only limited computational resources, to provide enrich context-aware services, e.g., location-aware user assistant services, from IoT environments, such services should be offloaded to be executed on server-sides, including cloud computing platforms. However, there are differences between access control models in context-aware services and cloud computing platforms, where the former needs context-aware access models and the latter widely uses role/subject-based access control models. This paper aims to bridging the models. We present a model for spatially specifying containment relationships of persons, physical entities, spaces, and computers to specify contextual information about the real world. Our approach connects between the world model and services offloaded to cloud computing as an access control mechanism. This paper presents the basic notion of the model and its prototype implementation.

Microservice architectural style emerged as a way of building highly scalable and flexible systems as opposed to the standard monolith approach. Despite the multiple benefits, as the number of services increase, the cost of service management and support also raises. In this paper we propose reference architecture for microservice systems in order to find a solution to the problem. The architecture approach is based on the notion of autonomic computing. It allows services to register or search for self-adaptation mechanisms when they need to respond to external environment changes.

The Internet of Things is a revolutionary concept, within cyberphysical systems, rich in potential as well as in multifacet requirements and development issues. To properly address them and to fully support IoT systems development, Agent-Based Computing represents a suitable and effective modeling, programming, simulation paradigm. As matter of facts, agent metaphors, concepts, techniques, methods and tools have been widely exploited to develop IoT systems. Main contemporary contributions in this direction are surveyed and reported in this work.

This paper presents our experience in integrating agents and robotics in our Computer Science Curriculum. We present a series of modules throughout our curriculum that progressively address these themes and other AI related topics, which ends with a specialised final year module central to teaching and learning multi-agent systems and principles of robotics. As part of this module a Robotics Challenge is organised, allowing students to integrate the knowledge they obtained in previously attended modules, and to practically apply knowledge and skills in order to solve a real problem.