Professional Statement of Ralf Moeller

Intelligent Autonomous Systems that we investigate are software systems based on representation formalisms such as:

• Description logics
• Probabilistic logics (e.g., Markov Logic)
• Temporal logics
• Logic programming (e.g., Datalog and Answer set programming)
• Functional programming
• ...

Intelligent systems in our view are computational agents

• that act in a locally rational way
• inside a designed mechanism
• such that a set of global goals is optimally reached (or reached with high probability).

Autonomous sytems in our context are software systems

• being responsible for a set of tasks
• that they carry out without permanent external control
• by receiving data from and acting on an environment,
• thereby adapting their behavior to optimally perform the tasks assigned to them.

Database systems and information systems in this spirit are

• no longer passive repositories simply used for standard fixed-index relational query answering and information retrieval but
• are designed as autonomic systems adapting to data profiles and query loads using machine learning techniques.
• They are modeled as systems proactively acquiring new information,
• possibly on the basis of goal-based information requirements, and by analysing data streams.
• New generation databases (semantic webs) support indefinite information and
• allow for query answering w.r.t. changing domain models (ontologies) and various ranking preferences.
• They also give feedback about data potentially missing and are able to summarize data sets as well as result sets.

Our work up to now has concentrated on query answering w.r.t. ontologies for large persistent data (see the dissertation Sebastian Wandelt). Expressive query languages for very expressive ontologies are investigated in Michael Wessel's dissertion. In addition, we have developed a system for the ontology-based interpretation of data streams (see also the dissertations of Atila Kaya and Sofia Espinosa). Interpretations are formalized as explanations computed using abductive reasoning on logic programming rules and are made practical by employing probabilistic ranking of abduction results (or explanations) (see the forthcoming dissertation of Anahita Nafissi).

Principles of model checking

• are required to develop basic data processing modules for autonomous systems
• and are the basis for reliable software systems in industrial applications.

The development of inference systems could certainly benefit from model-checking based verification techniques.
However, for, e.g., data intensive deductive query answering algorithms very few results on model checking are available.
Some initial results on certified model-based software development are presented in Miguel Garcia's dissertation.

Software systems and application areas we investigate are:

• RacerPro description logic reasoning system
• Cooperative Autonomous Systems such as UAV Swarms
• Logic-based Data Stream Interpretation (e.g. for High-level Logic-based Computer Vision)
• Inteligent Software Agents for the Semantic Web
• Machine-Learning Approaches for a Technical Analysis of Financial Markets
• Multisensor System Monitoring and Situation Recognition

Interested students are recommended to enroll in my lectures (listed here in alphabetical order):

• Architecture and Implementation of Database Systems
• Automata Theory and Formal Languages
• Computational Logic
• Foundations of Machine Learning and Data Mining
• Introduction to Database Systems
• Intelligent Autonomous Agents
• Multimedia Information Extraction and Retrieval

Relevant work can be found also the following resources (the list is definitely incomplete):

• Algorithmic Game Theory
• Autonomic Computing
• Autonomic Databases
• Autonomous Software Agents
• Autonomy Oriented Computing (AOC)
• Graph Theoretic Methods in Multiagent Networks
• Logical Analysis of Hybrid Systems
• Markov Logic
• Multiagent Systems
• Multi-Agent Systems
• Reasoning about Knowledge
• Reasoning about Uncertainty

Ralf Moeller