•The materialization of a Digital Twin(DT) ecosystem to enhance operational efficiency.•Utilizing the envisaged DT to employ different input space partitioning methods.•A novel RPM prediction method based on a reduced feature set and B-Splines.•A FOC model that combines the predicted RPM with vessel-specific characteristics.•Proposing ways to empower deep sequential or recurrent models with Splines.

Estimating the Fuel Oil Consumption (FOC) of a vessel is a critical task for the maritime industry, affecting route planning and the overall management of the vessel's operation and maintenance. Consumption is strongly coupled with the operation of the Main Engine (ME), but also with the environmental conditions (i.e., weather, ocean-energy spectrum) and the hydrodynamic features (i.e., resistance, propulsion) of the vessel. Current research shows that a multitude of features collected either from the AIS (Automatic Identification System) or on-board sensors can assist to the continuous prediction of FOC. Even when a FOC estimation model is perfectly trained on a specific vessel, its performance may degrade over time, when new weather conditions apply or when the hydrodynamics of the vessel change over time, due to fouling, aging and negligent maintenance. This work presents an online learning framework that employs a custom encoding-decoding Neural Network scheme and real-time data from various on-board sensors, to appropriately update FOC estimation models. The model is able to adapt to newly acquired data using a temporally-aware batch scheme, that samples from the initial training set using a custom auto-encoder.

Route optimization has been a research topic for many years in the maritime industry and it constitutes one of the key components to improving energy efficiency and sustainability in ship operations. This paper deals with the challenge of estimating Fuel Oil Consumption (FOC) in the context of Weather Routing (WR). Given a plethora of features collected from the vessel’s Automatic Identification System (AIS) or on-board sensor installations, we examine how a predictive FOC scheme can be coupled with WR optimization algorithms in order to reduce the vessel’s FOC, emissions, and the overall cost of a voyage. In order to handle the amount of data required for FOC prediction, we employ a streaming pipeline that harvests data in real-time from different sources and processes them appropriately for visualization, causal analysis, and forecasting purposes. In this direction, we first conduct an exploratory analysis to examine and unveil the importance and inter-association between the various variables related to sea-keeping and weather features, in order to utilize them effectively in the context of a FOC predictive scheme. Furthermore, we introduce a novel recurrent neural network architecture that approximates ideally the underlying function describing the features and the vessel’s FOC by taking into account historical data, and we showcase the results. Finally, we demonstrate how the FOC prediction model can be coupled with a WR algorithm to propose the optimal route for a vessel in terms of FOC efficiency.

In this work we face the challenge of estimating a ship's main-engine rotational speed from vessel data series, in the context of sea vessel route optimization. To this end, we study the value of different vessel data types as predictors of the engine rotational speed. As a result, we utilize speed data under a time-series view and examine how extracting locally-aware prediction models affects the learning performance. We apply two different approaches: the first utilizes clustering as a pre-processing step to the creation of many local models; the second builds upon splines to predict the target value. Given the above, we show that clustering can improve performance and demonstrate how the number of clusters affects the outcome. We also show that splines perform in a promising manner, but do not clearly outperform other methods. On the other hand, we show that spline regression combined with a Delaunay partitioning offers most competitive results.

Internet of Things (IoT) technologies have evolved rapidly during the last decade, and many architecture types have been proposed for distributed and interconnected systems. However, most systems are implemented following fragmented approaches for specific application domains, introducing difficulties in providing unified solutions. However, the unification of solutions is an important feature from an IoT perspective. In this paper, we present an IoT platform that supports multiple application layer communication protocols (Representational State Transfer (REST)/HyperText Transfer Protocol (HTTP), Message Queuing Telemetry Transport (MQTT), Advanced Message Queuing Protocol (AMQP), Constrained Application Protocol (CoAP), and Websockets) and that is composed of open-source frameworks (RabbitMQ, Ponte, OM2M, and RDF4J). We have explored a back-end system that interoperates with the various frameworks and offers a single approach for user-access control on IoT data streams and micro-services. The proposed platform is evaluated using its containerized version, being easily deployable on the vast majority of modern computing infrastructures. Its design promotes service reusability and follows a marketplace architecture, so that the creation of interoperable IoT ecosystems with active contributors is enabled. All the platform’s features are analyzed, and we discuss the results of experiments, with the multiple communication protocols being tested when used interchangeably for transferring data. Developing unified solutions using such a platform is of interest to users and developers as they can test and evaluate local instances or even complex applications composed of their own IoT resources before releasing a production version to the marketplace.

The need for low-cost health monitoring is increasing with the continuous increase of the elderly population. In this context, unobtrusive audiovisual monitoring methods can be of great importance. More particularly, the diameter of the pupil is a valuable source of information, since, apart from pathological cases, it can reveal the emotional state, the fatigue and the ageing. To allow for unobtrusive monitoring to gain acceptance, one should seek for efficient methods of monitoring using common low-cost hardware. This paper describes a method for monitoring pupil sizes using a common, low-cost web camera in real time. The proposed approach detects the face and the eyes area at first stage. Subsequently, optimal iris and sclera location and radius, modeled as ellipses, are found using efficient spatial filtering. As a final step, the pupil center and radius is estimated by optimal filtering within the area of the iris. Experimental results show both the efficiency and the effectiveness of our approach.

Recent advancements in single board computers, communications technologies and protocols, as well as the concepts of service-oriented architectures (SoA) and everything as a service (EaaS), constitute a prelude to the Internet of Things (IoT) revolution. Billions of devices are interconnected and integrated as modular web services, which can be used and re-used by developers making the building and realization of complex applications easier. In this work, we take advantage of the SYNAISTHISI platform, which is able to interface and integrate devices, services and humans, and expose their capabilities as virtualized semantically annotated services that can be mashed into applications. We analyze the development process from a developer's perspective, present an ontology for smart meeting rooms and focus on a real-world case, that is delivering a complex application for counting the persons in the interior of a smart meeting room, using technologies that support IoT.

This paper presents a method towards estimating a clinical depression-specific score, namely the Beck Depression Inventory (BDI) score, based on analysis of mid-term audio features. A combination of support vector machines and semi-supervised learning has been applied to map the mid-term features to the BDI score. The method has been evaluated on the AVEC 2013 depression dataset. The overall system has been implemented in Python achieving a 20 x realtime computational complexity on an average computer.

Unobtrusive every day health monitoring can be of important use for the elderly population. In particular, pupil size may be a valuable source of information, since, apart from pathological cases, it can reveal the emotional state, the fatigue and the ageing. To allow for unobtrusive monitoring to gain acceptance, one should seek for efficient methods of monitoring using com- mon low-cost hardware. This paper describes a method for monitoring pupil sizes using a common web camera in real time. Our method works by first detecting the face and the eyes area. Subsequently, optimal iris and sclera location and radius, modelled as ellipses, are found using efficient filtering. Finally, the pupil center and radius is estimated by optimal filtering within the area of the iris. Experimental result show both the efficiency and the effectiveness of our approach.

AI Planning and/or Scheduling techniques have been applied to problems that concern managing patient tests in hospital laboratories. In this paper, we propose the PTTPS model (Patient Test Timeline-based Planning and Scheduling) for planning and scheduling of patient tests in hospitals. PTTPS is an action-based model, which is based on the Constraint Network on Timelines (CNT) framework. The PTTPS knowledge representation schema enables the representation of various attributes of laboratories, equipment, hospital personnel, patients, tests, requests for tests, medical rules, as well as possible conflicts, constraints and appropriate sequence of tests. Patient tests are encoded as macro actions, each of which includes a set of subactions that represent steps of a patient test. We use CNT framework to produce appropriate constraints for the proposed PTTPS action-based model and we model an example of a patient test problem, the pyelography test. Keywords: AI planning and scheduling, constraint-based modeling, timelines, patient tests;

The BOEMIE project proposes a bootstrapping approach to knowledge acquisition, which uses multimedia ontologies for fused extraction of semantics from multiple modalities, and feeds back the extracted information, aiming to automate the ontology evolution process.

SUMMARY This paper presents the Web Usage Mining system KOINOTITES, which uses data mining techniques for the construction of user communities on the Web. User communities model groups of visitors in a Web site, who have similar interests and navigational behaviour. We present the architecture of the system and the results that we obtained in a real Web site.

In this paper we describe a semi-automated approach for ontology learning. Exploiting an ontology-based multimodal information extraction system, the ontology learning subsystem accumulates documents that are insufficiently analysed and through clustering proposes new concepts, relations and interpretation rules to be added to the ontology.

The basic goal of human robot interaction is to establish an effective communication between the two parties. In particular, robot emotion, speech, and facial expressions determine the way humans regard the robot, and they are deemed as essential for a natural form of communication. Addressing those issues is the focal point of this paper, while as a testbed we deployed a robot in a museum, where it serves as a guide to visitors. Our aim is a system that exhibits basic rational and intelligent behaviour.

We present the dialogue system of a robot that has been develo p d t serve as a museum guide. The robot interacts with human visitors in natural language, receiving instructions and p roviding information about the exhibits. Moreover, being m obile, it physically approaches the exhibits it provides information about. Alt hough the robotic platform contains many modules, includin g navigation, speech recognition and synthesis, our focus in this paper is the dialogue system, which supports the sessions between hu mans and the robot, as well as the natural language generation engine, wh ich generates the text to be spoken. Both modules are closely int rwined and depend on an ontology represented in OWL. The robot supports dialogues in both English and Greek.