The main topics of the presented papers focus on various aspects of maritime operations and security, including anomaly detection in maritime traffic, collision risk assessment, and the use of Automatic Identification System (AIS) data for enhancing maritime safety and surveillance. These papers cover a wide range of subjects within the maritime domain, such as trajectory clustering, kinematic behaviour analysis, Bayesian networks for risk assessment, resilience analysis of shipping networks, and the development of novel methods for detecting abnormal maritime behaviour. The emphasis is on using data-driven approaches, statistical methodologies, and advanced technologies to improve maritime operations and security.

Assistance systems play an important role in the proceeding transition of surface vessels towards highly automated operations. Particularly when navigating through congested areas like harbors, exact knowledge of distances to nearby obstacles is essential for collision avoidance. This paper applies a combined filtering and clustering approach in order to utilize 1D FMCW radar data for distance estimation to nearby obstacles in the harbor environment. The data processing aims at clearing the raw sensor data from unwanted signals caused by environmental influences like rain or waves and determines a reliable distance from the relevant signals. We evaluate our approach using sea trial data from a research vessel, comparing processed radar distances with a DGPS-based ground truth. The study assesses the performance of three density-based clustering algorithms-DBSCAN, HDBSCAN, and OPTICS-in this context. All of these algorithms show a good performance for processing the 1D FMCW data for our use case, enabling a reliable distance determination to a static obstacle. OPTICS performs slightly better in terms of eliminating disturbing signals than the remaining two algorithms. The processing times of all algorithms were found to be sufficient for online application of the proposed approach.

One of the major challenges that the drivers are facing today is increasing amount of information and multi-displays in vehicles. It results in drivers’ distraction, over workload and bad performance in driver-vehicle interaction, which may decrease the acceptance of novel technologies on autonomous vehicles. In this paper, based on the in-vehicle information systems (IVIS), we analyzed the in-vehicle dynamic information. We focused on Adaptive Cruise Control (ACC) using in the cutting-in driving scenario as a case study. Optimized HMIs on the head-up display and dashboard integrating basic vehicle information i.e., speed and ACC were designed. We proposed two HMIs: in HMI1, there was a temporal change, which means the detection information on ACC was displayed earlier; In HMI2, redundant information is spatially considered and distributed. Both HMIs were tested against a BASE HMI design, which displays all ACC-related information on the dashboard and the HUD. In a driving simulator study with a total of 30 participants, each participant performed in cutting-in scenarios with all three designs followed by a subjective workload assessment. The drivers’ lateral control behavior driving in the ego-car was measured. The results indicated that participants’ workload with HMI1 has a significant decrease compared to base HMI, while HMI2 did not show a significant difference. Neither HMI1 nor HMI2 affected driver behavior compared to the base HMI.

Prior to a voyage, a berth to berth planning is required to ensure safe sailing and also to have the autopilot setup correctly. During a voyage traffic supervision is one of the most safety critical tasks of navigators. Behavior prediction of other vessels based on few information and experience consumes a substantial amount of workload. The continuous increasing maritime traffic makes supervision a challenging task. This contribution elaborates a monitoring interface for maritime traffic supervision based on an Abstraction Hierarchy (AH) as part of an Ecological Interface Design process and compares the AH-driven design improvements with a mathematically derived solution-space design, originally targeted to air traffic observation, and with the current maritime standard HMI, the Electronic Chart Display and Information System (ECIDS). With the EID design subjects were on average more accurate and faster in identifying an overall critical situation and also more accurate in correctly identifying the most critical vessel if compared to the current state of the art (ECDIS) design.

With more goods to be transported oversea, traffic and vessels’ dimensions increase while berthing areas merely remain constant and thus challenge ship masters and pilots to maneuver in small basins with dense traffic even in bad weather situations. Too fast approaches or steep angles of attack result in damages to fenders, quay walls, or even impact the hull structure. We propose a shore-based, vessel-independent berthing assistant system to support sailors by Reference Points that are aligned to a quay’s meter markings and identify the precise berthing location by measuring distance and approach speed. For this purpose, we define the concept of a Berthing Support Area (BSA), which specifies an area in which, subject to constraints, safe berthing is provided. Within this area there are Reference Points, perpendicular distance measurements at arbitrary positions, which are implemented with a set of LiDAR sensors that have been integrated into the quay wall. In a test campaign with a vessel equipped with DGPS sensors, we sailed seven different maneuvers and evaluated the precision and the accuracy of the Reference Points for speed and distance measurements.

In harbor navigation and berthing, maritime pilots are facing today many challenges such as high dense traffic, changing environmental conditions and a lack of accurate information. In order to better handle this situation, systems are needed that offer improved situation awareness. This paper presents an Augmented Reality design concept for Smart Glasses to support maritime pilots in berthing and port navigation. Unlike other approaches, the extended docking support does not depend on ship-specific sensors, but benefits from a land-based infrastructure. To design the augmented berthing support for maritime pilots, the Konect method is applied. Finally, preliminary designs which serve as a basis for further research are provided in this paper.

The transformation of maritime navigation and control systems into an integrated System of Systems (SoS) consisting of a heterogeneous mixture of individual software-intense and safety-critical subsystems poses new challenges for the verification and validation of the overall system composition. Other than in traditional maritime architectures, the software-reliant structure of a SoS can be subject to change while already in operation, as features can be updated, errors can be fixed, or processes can be optimized. Thus, the alteration of a module on the system level necessitates the reassessment of compliance with the corresponding certification records. In this work, we present an approach on how the modules of a SoS can be associated through extended safety contracts with the corresponding safety case specification to verify the impact of a modification before deployment. Moreover, for each type of update (perfective, corrective, adaptive), the elements that need to be reassessed on the associated safety case are identified. Finally, the concept is established on a safety-critical module of the Maritime Traffic Alert and Collision Avoidance System (MTCAS) in order to assess the applicability of the developed approach.

The Automatic Identification System is a self-reporting system used by vessels and was introduced to enhance the operational picture on ship bridges. The Automatic Identification System destination port setting contains relevant information to anticipate a vessel's path. In future mixed traffic situations, autonomous vessels depend on correct destination port information specifically of human-operated ships to prevent dangerous encounter situations. In our Automatic Identification System data recordings of the last three months of 2018 a total of 4.988 unique vessels passing the German Bight with 13,216 different destinations were found. We found that at least 52.2% of all vessel destination settings are erroneous and a total of 1.3% (172) of the destination field settings were entirely conforming to the IMO UN/LOCODE recommendations. Our sample data indicates that no improvement in the percentage of correct destination settings has been made. Different to earlier studies, we report and quantify all eight error categories that we found and propose an algorithm that automatically adjusts the destination field settings. From those destination settings that two humans were independently of each other able to correct just by consulting a port and offshore dictionary (77,1%) the algorithm was able to correct 53,38% of the messages.

Vessels are getting more and more equipped with highly-automated assistant systems that benefit from the use of machine learning. Such trained safety-critical systems demand for new means of Verification and Validation (V+V). Their complex decision making process is hidden and traditional system analysis and functional testing is no longer possible as the testing space becomes too large to test. Scenario-based V+V performed in a simulation environment is a promising approach to tackle these challenges, triggering potential system malfunctions and covering as much as possible of the problem space. The authors propose a data-driven method to identify relevant sceneries, which describe states of a system in a scenario by a set of parameters. These states are derived from accident reports, summarizing the most critical situations a vessel and its automated assistant systems might be confronted with. By a chain of several methods, such as Principal Component Analysis and K-Mean Clustering the authors show that the value space of scenery parameters to be tested can be reduced and clusters can be identified that define equivalence classes of accidents. These clusters can then be partitioned depending on their probability distributions and open up a (reduced) space for random sampling of testing sceneries. The authors tested the method focusing on a weather-related parameter set of 1700 accidents in 2016 and 2017 that were retrieved from three different sources. Results show, that the first three principal components of the environmental parameters explain over 90% of the original variance and can be divided into 13 clusters. The authors then manually identified those accidents of a different data pool from 2013–2015 for that weather conditions were reported as the main cause of the accident and found the majority of them (61%) within the clusters and further 23% already in close distance. The more accidents are considered as input for the method the better would be the cluster fitting.

With the rise of electric and connected vehicles and an increased number of automated functions and sensors, the information shown to the driver changes. This contribution presents an engineering approach for designing and optimizing HMI interfaces to enable the driver to quickly be aware of all relevant information. In a study 6 master students in their final year of an HMI automotive design course either applied the engineering approach or used their preferred mix of methods to design a control panel that considers a pre-defined set of information to support three basic tasks: navigation, comfort and energy status. In a subsequent lab study, 27 students were presented with different situations on the designs. Participants should choose an action appropriate to the situation as soon as possible. The response time for the designs created with the engineering approach was 387 ms faster (p=0.015).

A human operator monitoring a safety-critical system has to gather information fast and accurate to detect problems and execute countermeasures in time. So far testing such HMIs is a complex task, since it requires HMI design prototypes embedded into simulated environments to perform tests with professional operators. We propose Konect Value, a heuristic to estimate the relative perception accuracy and operator reaction time already in the HMI design phase. The model-based estimation heuristic solely requires a task model and HMI design sketches as an input. The evaluation metric was applied to seven different HMIs, which were designed by Human Factor experts to support truck platooning. A comparison of the estimated accuracy and reaction times of Konect Value to a lab study (n=33) revealed high correlations for the relative reaction time (r=0.83, p<0.05) and also the relative perception accuracy (r=-0.90, p<0.01). This indicates that Konect Value is a promising heuristic for early HMI design evaluation in the safety-critical system domain.

A human operator monitoring a safety-critical system has to perceive information quickly and accurately to detect critical system states and execute countermeasures in time. So far, testing such human machine interfaces (HMIs) is a complex task as HMI design prototypes have to be implemented for simulation environments to perform tests with professional operators. We propose Konect Value, a quantitative method to estimate the relative perception accuracy and operator reaction time at an early design stage. The model-based method solely requires a task model and HMI design sketches as input. To validate the Konect Value, we applied the quantitative measure to seven different HMIs in a truck platooning use case. A comparison of the calculated value to the measured accuracy and reaction times in a lab study (n=33) revealed high correlations for the relative reaction time (r=0.83, p<0.05) and relative perception accuracy (r=-0.90, p<0.01). This indicates that Konect Value is a promising method for early HMI design evaluation in the safety-critical system domain.

The amount of information a human has to process continuously increases. In this regard, successful human performance depends on the ability of a human to perceive a system state as quickly and accurately as possible - ideally with a single glance. This becomes even more important in case several tasks have to be performed in parallel. It was shown earlier that monitoring user interfaces with a limited amount of information can be optimized for fast and accurate perception by combining all information into one integrated visual form. But systems that consist of several parallel tasks, each involving a whole bunch of parameters cannot be condensed into one single visual form. We propose an improved method that supports optimizing entire user interfaces consisting of several parallel tasks for fast and accurate perception (Konect). We evaluated the method in 6 workshops for that a total of 12 designers applied the method, which they learned by written instruction cards. Working in teams of two they were all able to design and optimize their designs first on a single task level (i.e. the original method) and thereafter on the global level (i.e. applying the new version). We evaluated their design outcomes thereafter in a laboratory experiment with 18 participants that were asked to distinguish critical and non-critical situations as fast and accurate as possible. Subjects were significantly faster ( $$p<0.001$$ ) and also significantly more accurate ( $$p<0.001$$ ) for those designs that were gained by the new version of Konect than those for the old one.

The consideration of driver's visual attention for Human Machine Interface (HMI) design is critical to ensure fast reaction times in unexpected situations and to promote situation awareness in hand-over situations. The effect of an HMI to the attention distribution of the driver can be measured by performing eye-tracking studies in a driving simulator. Performing eye-tracking studies requires functional HMI prototypes but give no insights on the underlying mechanisms for the measured behavior. In the tutorial we introduce a tool-driven and model-based approach to visual attention prediction, which can be performed already based on early HMI mockup ideas and with less effort compared to eye-tracking studies. The tutorial starts with an introduction to the theories of model-based visual attention prediction. Thereafter, participants are invited to either predict the visual attention for their own HMI design ideas or conduct an evaluation of an exemplary use case with the software tools that the participants can install on their computers or use in our lab.

The Human Efficiency Evaluator (HEE) is a model-based tool that predicts car drivers' visual attention based on a variant of the SEEV model. Different to prior research that required individual human factor (HF) expertise to generate valid attention predictions, the HEE enables to collect data from a group of experienced car drivers, to simulate human monitoring behavior, and to end up with valid predictions. We invited two different groups: automotive human factors experts (n=9) and experienced car drivers (n=20) to predict car drivers' monitoring behavior for a highway overtaking scenario with the HEE. Previous research did not detail the amount and experience of the HF experts involved in generating predictions, whereas our study revealed a quite high variance of individual HF experts' predictions about drivers' typical monitoring behavior. We measured car drivers' monitoring behavior using an eye tracking device in a car driving simulator (n=20). The aggregated prediction of the group of car drivers was high (R=0.719) and better than the average prediction of an individual HF experts.

Automotive HMI design is driven by systematic model based user engineering methods focusing on traceability and functional validation. User-centered design processes can open up the design space to discover new creative design solutions whereas model-based engineering methods offer a rigorous design derivation process.We present Konect, a user-centered design derivation process that fosters creativity and also ensures a systematic design derivation to end up with new interface designs that are optimized for correct and fast readability. Five Human Factor Experts applied the method and ended up with 5 creative and very different design results. In a follow-up experiment with 33 car drivers, we figured out that all designs could be perceived faster and more frequently correct than automotive interfaces for the same assistant system that have been designed with other design methods.

Automotive HMI design evaluation methods, such as usability assessments, attention and reaction time measurements require full working HMI prototypes to assess the usability and the subjects' performances in realistic situations. The theater-system technique and model-based prediction methods do not depend on functional HMI implementations and therefore promise HMI evaluation already in an early design phase. We applied both methods to evaluate three HMI designs for an Urban Adaptive Cruise Control (ACC) System. In a qualitative study with twelve participants, we used a theater-system-based technique to let them experience the HMIs in realistic situations. Subjects clearly preferred the HMI variant, which offers the best understanding of the vehicle's automation. By following a model-based approach, we evaluated the impacts to the driver's visual attention distribution of the three HMI variants with six human factor experts and found significant attention changes for the front window and for the Urban ACC HMI.

Monitoring is one of the most important tasks for an operator of a complex safety-critical system like a ship bridge or air traffic control. It is a prerequisite for good situation awareness. Designing an interface for such environments requires optimizing what is presented to the most limited resource: the operator's visual attention. But the real operator's attention distribution is hard to anticipate for a designer. We apply cognitive attention prediction methods to predict attention based on information gained on the one hand by the HMI designer and on the other hand the future user of an HMI. This contribution proposes and evaluates a set of comparative visualizations that support elaborating the differences between what has been designed and how it is perceived. An initial study indicated that a visually supported comparative analysis supports a designer in identifying differences and also seems to stimulate the designer to reason about the design.

Methods to get insights about users' monitoring behavior either depend on the expertise of Human Factor experts to model and predict stereotypic monitoring behavior or on performing eye tracking studies in simulated environments, which require subjects to be physically present and usually to be tested successively. AM-DCT is a tool that can be applied by domain experts without expertise in human factors and with limited training in parallel sessions to learn about a population's monitoring behavior. In an experiment 20 car drivers used the AM-DCT independently after watching a 15 minutes video tutorial. 19 subjects were able to model their monitoring behavior for a car overtaking scenario in 36 minutes on average. The identification of areas of interest for areas with clearly defined borders was very consistent among subjects. For those without clear borders an aggregated model of all participants seems surprisingly accurate to represent the real monitoring area.