Our group is involved in several national and international research projects, funded by the German Research Foundation (DFG), the European Commission (EU), the Australian Research Council (ARC) and others. For a list of finished research projects, see here.
 

The KI-LAB-ITSE is a joint initative of the six HPI research groups of Bert Arnrich, Jürgen Döllner, Tobias Friedrich, Robert Hirschfeld, Christoph Lippert and Christoph Meinel. The focus is on the topics of planning, developing and optimization of AI applications and AI systems. The chair for algorithm engineering contributes with research work on the foundations of artificial intelligence, e.g. in satisfiability and network algorithms.

Traffic congestion is an increasingly important issue that leads to longer travel times for everyone involved. It can be tackled by strategic routing, where (re)routing recommendations are distributed by traffic authorities and taken into account by the drivers' navigation systems. This central coordination will lead to faster, greener and safer traffic. In corporation with TomTom, students work on new algorithmic approaches for efficiently computing such strategic routes.

Boolean logic is the standard language to describe industrial decision problems stemming from various domains like logistics and optimization. Solving these problems corresponds to finding a satisfying assignment to the Boolean variables of a given formula. This can be found very efficiently with industrial solvers for instances with millions of variables. This practical efficiency stands in contrast to the most classical result of theoretical computer science: the proven computational hardness of propositional satisfiability (SAT). The disparity between these results is due to the stark structural differences between worst-case formulas and those that appear in practical applications. To address this, this project will study non-uniform and scale-free distributions, which appear to resemble industrial SAT instances more closely.

This project aims to build up and establish the area of evolutionary diversity optimisation. The project will cover the design and application of evolutionary diversity optimisation methods to complex problems of significance and high national economic benefit and build up the theoretical foundations of these methods. The project is expected benefit decision makers by providing them a diverse set of high quality alternatives to choose from. This project will allow them to make highly informed decisions and lead to more reliable solutions for optimisation problems, in areas of high economic impact such as manufacturing and supply chain management.

The project aims at improving the reliability of turbomachinery by incorporating physical substitute models. The developed virtual compressor is part of a system for permanent machine monitoring and predictive maintenance for large turbomachinery. The solution incorporates data acquisition directly at the machine, an IoT platform for saving and visualization of the data, and machine-dependent analysis tools. Our research group is responsible for developing suitable methods from artificial intelligence for anomaly detection and forecasts.

Network science is driven by the question which properties large real-world networks have and how we can exploit them algorithmically. Hyperbolic geometry has proven to be a useful tool in this regard. Our goal is to better understand the relationship between large real-world networks and hyperbolic geometry. We want to study properties of hyperbolic random graphs as well as embeddings of real-world networks into hyperbolic space. Moreover, we want to exploit these structural insights by developing algorithms that perform provably well on hyperbolic random graphs and, in turn, empirically well on real-world networks.

This project brings together theoreticians and practitioners who are working in the field of nature-inspired search and optimization heuristics. Its goal is on one hand to make theoretical findings more accessible for researchers and engineers who are concerned with applications, on the other hand to make theoreticians aware of practical problems and the kind of questions that arise in practical applications.

This project aims at using mathematical methods to identify the theoretical foundations of the capabilities and limits of computational learning. It uses a framework for learning criteria which allows for general statements, informing about many learning criteria at once.

The rest of this page lists our finished research projects.
 

Despite a large body of theoretical studies on load balancing, most results do not match the complex and heterogenous structure of the underlying network and the corresponding load balancing tasks. This project bridges the gap between the well-studied homogenous setting and the theoretically much less understood heterogenous load balancing problems that occur in practice.

Time series data is data derived from consecutive measurements over time. It appears in all domains of applied science and engineering. Storing time series data can quickly overload any available storage. The random noise introduced by physical measurements fundamentally limits the achievable compression rate. We want to evaluate existing algorithms and develop new algorithms for lossy compression schemes for high-frequency time series data from different domains. The key question is what compression can be achieved without losing too much information while still being able to analyze the data and using it for modeling and statistical predictions.

Bio-inspired swarm algorithms are well established in practice for solving optimization problems with complex constraints even in difficult domains involving uncertainty. This project analyzes theoretically swarm-based search heuristics for dynamically changing and stochastic objectives.

The aim of this project is to design bio-inspired computing methods for dynamically changing environments. Dynamic problems arise frequently in the areas of engineering, logistics, and manufacturing. Such problems are usually subject to a large set of constraints that change over time due to changes in resources. Algorithms that can deal with such dynamic changes would benefit decision-makers. The project aims to provide a foundational theory as the basis for the design of bio-inspired algorithms dealing with dynamically changing constraints and provide approaches for dealing with important industrial problems.

All medium and large corporations are required to have their financial statements reviewed by an auditor. Each auditor has a certain set of skills and available time slots. Performing an audit requires different skills at different times. This results in a complex scenario with many hard and soft constraints to satisfy. In corporation with KPMG AG Wirtschaftsprüfungsgesellschaft, eight bachelor and three PhD students work on a new scheduling algorithm.

The Navigation Data Standard (NDS) is a new regularization for map data used, for example, in portable navigation devices. It restricts the data to be only accessible in blocks corresponding to the earth’s grid. Due to the limited memory of portable devices, it is not possible to store the entire map in memory at once. In corporation with TomTom, five students work on new algorithmic approaches for efficient routing algorithms that load as few blocks as possible.

Biological evolution has produced an extraordinary diversity of organisms. Evolutionary computation has found many innovative solutions to optimisation and design problems. Both fields have studied the speed of adaptation independently, and with orthogonal approaches. This project brings together an interdisciplinary consortium from both fields to synergise these complementary approaches and to create the foundation of a unified quantitative theory describing the speed of adaptation in both biological and artificial evolution.

This project establishes the field of parameterised analysis of bio-inspired computing. It rigorously analyses features of instances of combinatorial optimisation problems and their impact on the runtime behaviour of bio-inspired computing methods such as evolutionary algorithms and ant colony optimisation.  

Looking for a parking place in urban areas is often time consuming and stressful. Up-to-date navigation systems are barely capable of directing the driver to a good nearby parking lot. This is mostly due to missing data about available parking places on the roadside and little research on this topic. In corporation with TomTom five students work on new algorithmic approaches for efficient on-street parking.

This project aims to analyze the average-case behavior of some known parameterized algorithms, mainly motivated by the recent encouraging experimental studies of these algorithms. Most of these studies record, compared to the worst-case analysis, a much more promising performance on real-world data.

Two main approaches have been considered in dealing with NP-hard problems in the last decade. One is parameterized complexity theory. The other one follows a probabilistic way of analyzing problems and algorithms. This project aims at developing the necessary tools and theories to combine parametrized and smoothed complexity.

Swarm based randomized search heuristics such as ant colony optimization (ACO) and particle swarm optimization (PSO) are established in various applications and particularly in dynamic optimization problems, providing good solutions. Unlike e.g. in evolutionary algorithms (EAs), hardly any theoretical foundations existed before the project started. The main purpose of this research project was to analyze the efficiency of different approaches.