In this talk I will provide an overview of some of the work we do when it comes to solving inferenceproblems in nonlinear dynamical systems. A proper subtitle for the talk is "strategies andexamples", since I will only provide solution strategies and show that these strategies cansuccessfully solve challenging applications. The first topic is parameter inference problems innonlinear dynamical systems (a.k.a. nonlinear system identification). The maximum likelihood problemis solved using a combination of the expectation maximization (EM) algorithm and sequential MonteCarlo (SMC) methods (e.g., the particle filter and the particle smoother). The Bayesian problem issolved using a combination of Markov chain Monte Carlo (MCMC) and SMC. As an example, we show how toestimate a particular special case known as the Wiener model (a linear dynamical model followed by astatic nonlinearity). The second topic is that of sensor fusion, which refers to the problem ofinferring states (and possibly parameters) using measurements from several different, oftencomplementary, sensors. The strategy is explained and (perhaps more importantly) illustrated usingthree of the industrial applications we are working with; 1. Navigation of fighter aircraft (usinginertial sensors, radar and maps); 2. Indoor positioning of humans (using inertial sensors andmaps); 3. Indoor pose estimation of a human body (using inertial sensors and ultra-wideband).
Thomas B. Schön was born in Sweden on December 25, 1977. He is Associate Professor with the Divisionof Automatic Control at Linköping University (Linköping, Sweden). He received the PhD degree inAutomatic Control in Feb. 2006, the MSc degree in Applied Physics and Electrical Engineering inSep. 2001, the BSc degree in Business Administration and Economics in Jan. 2001, all from LinköpingUniversity. He has held visiting positions with the University of Cambridge (UK) and the Universityof Newcastle (Australia). He is a Senior member of the IEEE. He received the best teacher award atthe Institute of Technology, Linköping University in 2009.
Schön's main research interest is nonlinear inference problems, especially within the context ofdynamical systems, solved using probabilistic methods. He is active within the fields of machinelearning, signal processing and automatic control. He pursues both basic research and appliedresearch, where the latter is typically carried out in collaboration with industry. More informationabout his research can be found on his website: users.isy.liu.se/rt/schon/researchOverview.html