A domain-specific architecture framework for the maritime domain
Stakeholders in the maritime domain face the challenge to harmonize existing systems and integrate new approaches and technologies into existing technical and organizational structures for sustainable, reliable and safe maritime transportation. This affects technical components, organizational structures and human users as well as the common interaction as elements of socio-technical systems. Following IMO’s e-Navigation vision for seamless integration between existing and upcoming systems, existing systems have to be analyzed in a structured way on their system architectures including the specific context and organizational structures in which a system is embedded to enable views from different operational or technical perspectives on the examined systems. This will enhance the understanding of the relationship between the maritime world and the examined systems and to consider this during the development of new maritime systems. An architecture framework shall provide a consistent methodology to structure the engineering process of socio-technical system concepts and to align technical systems. Hence, such a framework needs to frame the maritime domain including its stakeholders, the existing and upcoming technical system (architectures), related business processes and organizational structures including governance and regulation aspects to enable a complete view on the maritime infrastructure.The scope of such a domain-specific framework is to structure the system engineering phases starting from planning over the identification of requirements to the use case development in a harmonized and formal way. This allows the user to map the results, to visualize them, explore interoperability issues between system architectures and to identify spots which need to be standardized.
Sensor Integration and Control of Marine Vehicles
In a marine vehicle control system, sensors data include body positions, orientation and trajectory, and the ocean environment, human reports, etc. Therefore, sensor integration is a necessary technique to synthesize all the sensors data to provide more reliable and accurate information. Nevertheless, coupling, drifting and other errors are always included in the measured sensors data, the classic method to deal with the data is kalman filter or extended kalman filter estimation, and they result in estimates for the fused sensors data that are optimal in a statistical sense. However, there could still be some unexpected errors causing wrong behavior of estimated ship’s dynamic, so as a candidate solution, fault-tolerant control can detect and prevent those unacceptable errors and limit the remaining errors within a safe range, in order to control the vehicles with the desired motion on its predefined trajectory.
Damage Mitigation Techniques of ship collisions and grounding
(Hasan Youseff Deeb)
My research topic concerns the phase of encountering situation between two vessels, where collision is not avoidable anymore. it focuses on defining the main collision scenario variables and investigates the ability to use such collision scenario variables in order to minimize the expected damages. e.g. effects of collision angle, speed and collision location (structural distribution) on impact energy and damage volume. My research will investigate the ability to use ship maneuvering capabilities in order to modify such collision scenario variables to reduce collision impact energy i.e. Provide emergency steering guideline (Algorithm) to make evasive maneuvering more wiser once collision is inevitable. same applies for grounding accidents with known shallow water area.
Concepts for Spatio-Temporal Anomaly Detection in Data Streams
Spatio-temporal data describes data with additional information about location and time. The amount of spatio-temporal data increases rapidly, as mobile devices with satellite based navigation systems, such as GPS, are becoming smaller and widely available. Not only smartphones, but planes, ships or even animals can be equipped with location sensors. With an additional internet connection, they can send their position in near real time and can attach additional measurements, such as the temperature, destination or height.Such data, especially if it is send live via an internet connection and the amount of data is very high, is difficult to oversee by humans. Data mining on spatio-temporal data is therefore needed to find common patterns, detect anomalies or predict future developments. If the data is send live in near real time, the processing of the data should be done in near real-time as well to get fast results. In addition, the high amount of very detailed data may be too much to save everything in a database and analyze it afterwards. Therefore, data stream processing is a useful approach for such data. Here, not all data is stored permanently but the processing is done on a smaller, more relevant set of the data.This research topic aims to analyze, use and improve concepts for spatio-temporal anomaly detection on data streams. The concepts can for example be evaluated on AIS (Automatic Identification System) data from vessels, which contain the current position of the vessel as well as it's identification number and destination among others. The evaluation can be done within the Odysseus Data Stream Management System (see odysseus.offis.uni-oldenburg.de/).
Multi-Tenancy/Multi-User Architecture for Data Stream Management Systems in Maritime Applications – Resource Management, Privacy and Isolation Aspects
Description language for environment scenarios generation for maritime simulations
Environment scenarios in vessel simulation contains various environment factors that influence the movement of the vessels. These scenarios play crucial roles in multi-agent vessel simulations and usually are assumed to be given. However, in practice, they are not always available. Firstly, maritime experts desire to see different kinds of edge cases, such as how the ship move through the ocean under extreme weather conditions. Using the existing observation data or prediction data, it is unlikely to cover every cases. On the other hand, the experts may want to manipulate the behavior of the factors, e.g. speeding-up the wind at a stable frequency. Existing datasets does not provide this kind of freedom. It becomes time-costly when maritime experts have to switch their focus to the structure of spatial-temporal coverage data, trying to generating scenarios which satisfy their needs. This research aims at providing a formal description language together with its execution engine, to solve the above-mentioned problem. It frees the maritime experts from understanding how the spatial-temporal coverage data is organized and generated. The language grammar allows them describing the properties and behavior of the marine environment factors using the domain abstractions they are familiar with. The execution engine renders the scenario file to generate a corresponding spatial-temporal data frame, fills in the suitable values of required environment factors. By changing the description, the scenarios can be easily modified.
Ship Dynamic Modeling, and Identification Using Observation Data
To devise and implement a control strategy giving a vessel the ability to robustly and accuratedly predict trajectories of own and others that span the vessel's entire performance envelope. To fulfill this task a model which is sufficiently rich to enable effective motion planning and control and sufficiently simple to allow straight forward parameter identification much be selected from a set of simple models whose parameters can be quickly and easily identified from standard motion data. Having selected an appropriate model, dynamically feasible trajectories can be predicted by using AIS data for the own ship and other ships to take action for avoiding collision. An effective closed-loop control algorithm then enables predicting aggressive trajectories in the presence of environmental disturbances and modelling uncertainties.
Nonlinear Model Predictive Control for Path following and Collision Avoidance
My research problem is representation of path following and collision avoidance of vessels as a Multi agent dynamic game with an objective to optimize a coupled payoff function subject to the dynamics of our vessel and the nearby vessels. In order for that collision avoidance system to be efficient, the algorithm must predict other ships states to avoid any future possible collision in a finite predetermined horizon. So, the dynamic game will be modified to include the predicted states in the coupled cost function and solve the optimization problem at every sample in a moving horizon manner. This is called in the literature distributed model predictive control (DMPC). To guarantee Collision avoidance, the time delay in predicting other nearby vessels must be taken into consideration and not neglected.
Integrated nautical Maneuver planning for modern SHIPcontroller
Modern steering devices for ships lead to an improved control capability and allow independent actions without tugs by mates in use with modern electronics. Within the project AdaNav developed MIMO controller enables DP equivalent functions for ships being able to traverse. The dissertation will discribe in theorie and practice the inclusion of such modern maneuvering quality as computer based assistance system into electronic ship guidance systems like ECDIS. To guarantee fully responsibility of mates the concept bases on introduced planning practice implementing additional waypoints and attributes in a so-called kinematic sequence. This kinematic sequence connects the developed integrated nautical user interface and the existing MIMO controller of the AdaNav project.
Event detection for maritime time-series data with computational intelligence
The growing infrastructures of information and sensor systems in the marine domain lead to an increased demand of data mining solutions. Especially time series data from sensor platforms with high dimensionality create massive amounts of data daily. Possible application with such data are event or outlier detection and value prediction for forecasts. The problems of existing methods are that often data are handled if they are univariate or the time relation is ignored. Further, this real data is prone to become missing and having high noise rate, which results in a decrease in the usefulness of data. My first publication was about the marine event detection with data from the sensor platform at Spiekeroog. The applied machine learning algorithm was local outlier factors and tested three different preprocessing methods. Evaluation was conducted with help of domain experts, who classified found events as positives or false-positives. The results indicate that events of interest to domain experts can be found with an accuracy of 80% and the preprocessing is crucial for the success of event detection. The next step is repairing missing data to enable analysis of complete data. For example, I had to exclude a whole month from the event detection, because the algorithm could not process incomplete data. I utilize dynamic time warping and a simple interpolation step to address the specific problems of missing data from sensor platforms, which can be conclusively missing and multiple missing values at the same time. In future work I will use Machine Learning to perform predictions of values that are in another place of the ocean to augment sensor platforms with measurements they can’t conduct at this time, e.g. for the BEFmate project. This will feature multiple data sources at once from the Cosyna project. Also, more in depth detection of ocean events are planned.
Evaluation of Distributed Situation Awareness on a Ship Bridge
Situation Awareness (SA) is an important cognitive aspect of the decision making activities on a ship bridge. The navigation crew uses the bridge systems to gather information elements and gain SA. Designers need to understand the SA requirements of the crew and develop systems with user interfaces (UI) that support these requirements. The existing methods for evaluating SA are more fitting for assessing the SA of individuals rather than the Distributed Situation Awareness (DSA) that characterises the ship bridge environment. This paper proposes a model-based approach for the evaluation of DSA that can be used in early system design stages. The methodology will use fuzzy cognitive mapping techniques to represent the DSA requirements of the crew. The anticipated results include a cause-effect analysis tool and suggestions for UI design of intelligent systems.
Improving the situation awareness of OOWs by tracking and shifting their visual spatial attention
(Tim Claudius Stratmann)
With ships becoming bigger and bigger, navigation and maneuvering without the given technology has become nearly impossible. Seafarers have to rely on information given by their navigational and assistive systems. I am aiming to improve the situation awareness of officers of the watch by optimizing their monitoring behavior. To achieve this, I will track the visual spatial attention of the officer of the watch and shift it to areas with potentially unseen critical information. This should lower the risk of sea accidents such as collisions and grounding.
Be aware of your surrounding - Visualization of dynamic
Off-Screen objects and Attention Shift in head-mounted Augmented Reality
(Uwe Wilko Grünefeld)
Modern working environments are still challenging the human worker. This
is mostly due to widely distributed areas, because one person is unable
to focus all these relevant areas at the same time. An example for such
a working environment is the ship bridge of big container vessels. In
many different scenarios there are several objects to keep an eye on
like a vast number of displays, other vessels, buoys, tugboats and quay
walls. Compared to the limited human field of view and the even more
limited focusable area it gets clear, that it is impossible to have all
relevant objects in vision. In my work I want to develop a solution for
this problem based on head-mounted Augmented Reality. This solution
should be able to visualize objects out of vision (so called Off-Screen
objects) and if necessary shift the attention towards these objects. The
visual modality should be used, because auditive and tactile signals are
not fitting because of high density and complexity of the information.
For example many Off-Screen objects at the same time with distance and
type information which changing position frequently. Auditive is also
not used because it is already highly used for radio communication,
alarms and communication to other people on the bridge. Furthermore
head-mounted Augmented Reality is used because the normal working flow
shouldn’t be interrupted for example because the hands stay free and
because information is only addressed to one person specifically.
Development of a MIMO control framework as an universal navigation control
Modern vessels possess advanced actuation systems like azimuth or bow
thrusters to archive better control characteristics and energy efficiency. This
thesis includes an evaluation whether the necessary parameter setting for
automatic control of such actuators can be simplified, generalised or automated
using abstraction models to enable a flexible, reliable control system.
Towards enhanced capabilities for automatic detection of small, non‐metallic maritime artefacts with remote sensing radar data
The project aims at enhanced capabilities for automatic detection of small, non‐metallic artefacts at the ocean surface. Described project’s outcome should facilitate a fast, automatic localization of those special targets in a (theoretically) continuous available data stream of radar satellite images. That capability that could a central technical building bloc to support Search & Rescue missions and therefore facilitate fast and sustainable disaster mitigation. A web service visualizing and disseminating detection of targets in near real‐time could be an ostensive application, too. Yet another possible implementation could be enhanced detection of maritime plastic pollution. Experiencing substantial and active developments in the field of satellite technology and fast rising numbers of operable satellite systems, there is the prospect of continuous enhancement of spatial as well as temporal resolution of available remote sensing data in coming years. This perspective represents an essential fundament for further developments of analysis and processing methods and furthermore for a diversification of subsequent applications ‐ within this project and beyond.
Underwater acoustic system integration with maritime test-beds for ship maneuvering parameter extraction
Extending the service provided by maritime test beds with systems that remotely extract extra information from passing vessels without the need of direct cooperation is a potential advancement. The project aims to develop a new system that uses underwater acoustic technologies, and integrate it with available maritime systems. This integration enables the extraction of new set of information that serves several applications in the maritime domain. The maneuvering parameters such as propeller speed and rudder angle change play key role in ship behavior prediction application and environmental monitoring fields.