Dresden 2014 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

BP: Fachverband Biologische Physik

BP 27: Modelling of non-linear dynamics in biological movement (focus session, joint BP/DY)

BP 27.7: Talk

Wednesday, April 2, 2014, 15:45–16:00, ZEU 250

How to turn the non-linear muscle into a linear all-purpose tool — •Karl Theodor Kalveram — Heinrich Heine Universität Düsseldorf, Germany — Technische Universität Darmstadt, Germany

The three basic categories of biologically motivated tasks that we discriminate we call "reaching", "cycling" and "enforcing". Because in all these activities the physical environment has to be influenced in a scaled manner, the organism must provide appropriately scaled forces. Our muscular-skeletal system solves those problems. We ascribe this to the organism's property to generate forces by muscular activation, and to generate this activation through neural stimulation. It remains, however, the open question, how to specify that stimulation, which exactly produces that forces, which are necessary to complete the respective task correctly?

Here we propose a control schema, which makes the non-linear Hill-type muscle a multiple-purpose tool for solving the biologically imposed motor tasks mentioned above. We achieve this by training an artificial neural network by a two-step auto-imitative learning algorithm (a special type of learning by regression), which makes the network an adaptive inverse controller of the physical environment to be controlled.