Dr. Jussi Rintanen Aalto University Department of Information and Computer Science also: Adjunct Associate Professor, Institute for Integrated and Intelligent Systems, Griffith University, Queensland, Australia email rintanenj at gmail dot com

My research interests are in search and automated reasoning, mostly for solving discrete combinatorial problems. A main problem I have been interested in is in finding state, action, event or task sequences in very large transition systems. This problem appears as a subproblem when deciding what to do and how to do it, which can be viewed as a planning or controller synthesis problem. Most of the research challenges are caused by the astronomically high numbers of possible state/event/action/task sequences, and the alternative sequences that have to be considered when we cannot predict exactly what is going to happen. My research interests here are in the junction of the theoretical and the algorithmic: I am interested in understanding search methods such as SAT algorithms (propositional resolution proof systems) better, especially their capabilities in solving state-space reachability problems, as well as the capabilities of other symbolic (non-enumerative) search methods such as Binary Decision Diagrams, and the relations of these methods to enumerative state-space search algorithms. A recent addition to my interests is the solution of problems that combine discrete (Boolean) reasoning with probabilities, for solving various problems in probabilistic reasoning and machine learning.

In the application side, we are interested in discrete and hybrid systems control (planning), monitoring, and diagnosis, in applications such as the Smart Grid (intelligent electricity networks.) A further area in our research focus is cyber warfare, especially as a part of network-centric warfare, as well as coordinated response, contingency planning, and red teaming. We try to understand, automatically, what happens in a complex system (monitoring, diagnosis), what could happen (contingency analysis), and what actions to take to avoid problems and to recover from fault situations (control). Another application area for our research is business processes. We are especially interested in goal-driven and declarative business process modeling, synthesis and execution. The main problems in these applications can be understood in terms of (discrete or hybrid) state-transition models, and the a main goal of the research is to transfer the results of our research to solve the application problems better, and to drive the research by the problems arising in the applications.

I obtained my PhD in Computer Science from the Helsinki University of Technology in 1997, held research and teaching positions at the universities of Ulm and Freiburg between 1997 and 2005, and have been in Australia since January 2006. See my CV for details.

Research

• Factored search methods for state-space search problems, especially based on model-finding and logic-based data structures.
• Application of combinatorial methods to diagnosis, decision making, probabilistic reasoning and learning, and synthesis of plans, controllers and policies
• Representation of high-level system specifications, as needed for high-level synthesis of business processes, workflows and information systems.
• Supervisory control of discrete and hybrid systems, such as electricity networks and other utility infrastructure.

supervisory control of Smart Grids
automated planning and decision-making in military operations and cyber-warfare
program synthesis in knowledge-intensive information systems
planning as satisfiability (software, papers, links)

Presentation materials

Thesis topics

• scheduling for close-to-optimal snow-plowing The goal of the project is to produce a scheduler for finding good quality routes for fleets of snow-plowing equipment. First, a full-information scheduler is developed. Later, the scheduler is extended to handle uncertainty for example with respect to completion times and forecasts for snowfall.
• Weighted model-counting and #SAT Algorithms for the model-counting problem are investigated, in connection with important applications from probabilistic reasoning.
• Efficient implementations of algorithms for SAT by using (not completely pure) functional programming languages The project investigates the issues affecting the efficiency of high-performance combinatorial search algorithms, in this case for SAT, when implementation is done with a strongly typed functional language such as Standard ML. The development of CPU technology, especially the strong reliance on fast multi-level caches, has changed the overhead different programming languages have in comparison to machine code and low-level languages such as C. In the project, a high-performance SAT solver is implemented in Standard ML, with an experimental (and possibly analytical) investigation of the impact of implementation specifics on performance.

Photos: panoramas, birds