# Regensburg 2013 – wissenschaftliches Programm

## Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

# DY: Fachverband Dynamik und Statistische Physik

## DY 12: Nonlinear Dynamics, Synchronization and Chaos I

### DY 12.3: Vortrag

### Dienstag, 12. März 2013, 10:00–10:15, H48

**Estimating forces by observing a noise-driven mechanical system** — •Andreas Ruttor, Philipp Batz, and Manfred Opper — Technische Universität Berlin

Predicting the dynamics of a mechanical system is easy as long as its properties, e.g. moments of inertia and friction coefficients, are exactly known. But for real devices, e.g. robots, this is usually not the case. Applying a noisy control force leads in combination with friction to a stationary stochastic dynamics. We show that one can estimate the forces in the system as a function of its state by observing these movements. While it is possible to measure accelerations directly, this approach requires very dense and evenly spaced observations. In contrast, our method estimates the probability distribution in phase space, which works with larger time intervals between data points, too. It is even possible to choose the position of the observations randomly. Prior knowledge about the system can be included in a parametric model of the potential energy. But we also have a non-parametric approach based on Gaussian process regression, which works without that information.