I studied computer science with a focus on operations research and artificial intelligence. During my PHD studies, my area of expertise expanded to model-based reasoning and probabilistic reasoning. The former deals with methods to automatically derive conclusions about some system from a model of this system and observations from this system. Such a system could be, for example, a car, satellite, printer or a factory plant. The latter is concerned with methods to automatically arrive at conclusions in a probabilistic way, which regards the quantified uncertainty in the knowledge that forms the basis of such reasoning.
More specifically, I gained experience with logical and probabilistic modeling, constraint optimization, combinatorial optimization, model-based diagnosis and probabilistic inference. All this can be understood as "artificial intelligence", maybe in a broader interpretation of the term.
I did my PhD studies at Informatics/TUM in the group of Dr. Martin Sachenbacher.
In my thesis, I address a computational problem called plan assessment, which occurs when we try to design technical systems that operate autonomous to a certain degree. Such a system automatically creates plans to achieve given goals and then executes these plans. For example, a manufacturing plant may plan the production for individual products and then produce them. Since the variety of products, or in general goals, is large, successful plan execution is hard to guarantee beforehand. When facing contingencies, The system has to react autonomously. To do that, it needs to know how probable it is for each goal that it will be reached successfully. In the example, we want to know how probable the successful production still is for each product. And in case the probability is low, we also would like to know potential causes. In my thesis I develop approaches to compute answers to these questions. The approaches are unique in that they rely on two important ingredients: they use probabilistic models of systems that are formally specified in a way typical for engineering and they exploit powerful off-the-shelf implementations of generic algorithms developed in Artificial Intelligence.
Selected Publications
F. H. Petzschner P. Maier S. and Glasauer: Combining symbolic cues with sensory input and prior experience in an iterative bayesian framework. Frontiers in Integrative Neuroscience, 6(00058), (2012)
P. Maier and D. Jain and M. Sachenbacher: Compiling AI Engineering Models for Probabilistic Inference, In proceedings KI-2011,7006,191--203,Springer,(2011)
P. Maier, D. Jain, S. Waldherr and M. Sachenbacher: Plan Assessment for Autonomous Manufacturing as Bayesian Inference, In proceedings KI-2010,6359,263--271,Springer,(2010)
P. Maier, M. Sachenbacher, T. Rühr and L. Kuhn: Automated Plan Assessment in Cognitive Manufacturing, Adv. Eng. Informat.,24,241--376,(2010)
P. Maier and M. Sachenbacher: Diagnosis and Fault-Adaptive Control for Mechatronic Systems using Hybrid Constraint Automata. In proceedings of the conference of the prognostics and health management society PHM-2009, (San Diego, CA, USA, 2009)
P. Maier and M. Sachenbacher and T. Rühr and L. Kuhn: Integrating Model-based Diagnosis and Prognosis in Autonomous Production, In proceedings of the conference of the prognostics and health management society PHM-2009, (San Diego, CA, USA, 2009)
