Maps are one of the most efficient ways to communicate information. They help people to make decisions in navigation, spatial planning, or risk and disaster management. Maps also communicate geopolitical information, they give a spatial dimension to rhetoric arguments, and generally aid the process of public opinion and consensus building. Effective maps immediately convey their message and hence are as simple as possible. Schematization creates a simplified and compact representation of the original data and reduces the visual complexity of maps. Linear features, such as roads and rivers, and the boundaries of regions are often drawn using only a few straight line segments in few different directions, or they are approximated by a few simple curves. Traditionally schematized maps make extensive use of curves. Curves have greater expressive power than line segments, several line segments can often be replaced by a single arc of a low degree curve. Curves also make it easier for users to interpret maps. However, automated methods for schematization are mostly restricted to straight lines. Decision makers and the greater public benefit from high quality on-demand map production, which necessarily has to be fully automated. Hence we propose to develop algorithmic methods for automated curved schematization. We aim to study curved schematization for region outlines, subdivisions, and networks. We will establish cartographic quality criteria and develop efficient algorithms that take these into account. Formalizing these criteria and testing their validity will be an iterative process executed simultaneously with algorithm development and experimental evaluation.

The availability of devices that can be used for tracking objects has increased tremendously over the past years, leading to an explosive growth of data recording the trajectories of objects. To make effective use of these data, efficient algorithms and data structures are needed. Most research to date focuses on efficient algorithms for analyzing trajectories: computing the similarity between two given trajectories, clustering a set of trajectories, finding motion patterns, and so on. The data-structuring side of the problem---how can we preprocess and store a large collection of trajectories such that certain queries on the collection can be answered efficiently---has received much less attention. This is the goal of our project, where we will focus on various types of similarity queries: report (or count) the (sub)trajectories in the collection that are similar to a given query trajectory. This is highly relevant in applications, but also very challenging from a theoretical (computational-geometry) point of view, since there are hardly any data structures available for query problems where both the collection to be stored as well as the query consist of complex objects. Our goal is to develop solutions that work well in practice and whose performance is supported by a theoretical analysis of their speed and accuracy under reasonable assumptions on the input data. We will not only develop the theory of such data structures, but also provide prototype implementations and perform an experimental analysis.

EWAF aims to focus on advancing the computer science side of algorithmic fairness (novel fairness-aware AI techniques analyzed theoretically and empirically) and grounding it within the context of Europes legal and societal framework, particularly in light of the EUs efforts to promote ethical AI and entering into force of the EU AI Act. EWAF25 will build on the success of the first three editions. An increasing number of European researchers are now contributing to this domain. Of course, EWAF remains open to contributors and participants outside Europe as well.

Generating control software with graphics processors Many devices and systems are controlled by software. Think of elevators, printers, cars, planes, bridges, and assembly lines in factories. Correct control is often crucial for safety. Supervisory Controller Synthesis is a technique to automatically derive control software, making the software correct by construction. The involved calculations, however, currently limit its applicability. In this project, we conduct research on how graphics processors, with an enormous computational power, can be used to scale up this technique, making it possible to generate more complex software in a shorter time span.

The amounts of data available to business, industry, science, and governments have become too large to analyze them with automated techniques or visualization alone. Expert knowledge of a human analyst is required and should be integrated in multiple stages of the analysis process. Visual analytics is an emerging research area that aims to do exactly this by using interactive visualizations to integrate computational methods for data analysis with the knowledge and experience of the analyst. To draw conclusions from the analysis process, the analyst requires reliable output from the system. Therefore, the underlying algorithms should provide a guarantee on the quality of their results, a topic that has not received much attention in visual analytics research. The highly interactive visual analytics process bears unique challenges and requires a novel perspective on how to measure the performance of an algorithm. The aims of the project are to develop algorithms, which (i) give a fast response and then refine the solution progressively, (ii) provide the user with the means to steer the computation and are flexible enough to adapt to changing objectives, and (iii) provide a guarantee on the quality of their results. Also, we will design and implement appropriate and effective metaphors for presentation and novel mechanisms for interaction with the algorithms. The goal of this project is to build a framework for progressive, user-steered algorithms in visual analytics. Our main application area will be spatio-temporal data, in particular movement data, but we aim at generic solutions.