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

BP: Fachverband Biologische Physik

BP 37: Modelling of non-linear dynamics in biological movement (focus session)

BP 37.4: Vortrag

Mittwoch, 18. März 2015, 16:00–16:15, EW 202

Reafference Principle 2.0 — •Kim Joris Boström and Heiko Wagner — Motion Science, University of Münster, Germany

The reafference principle was introduced 1950 by Holst and Mittelstaedt, and its basic features have been confirmed by many experiments. It holds that the neural systems makes an efference copy that is compared with the reafference, i.e. the afferent signal resulting from movement caused by the efference, and the difference is passed to higher centers. However, efferent and afferent signals encode very different kinds of information, between which there need not exist a linear relationship. To address this problem it has been suggested that the brain involves a forward simulator to calculate the expected reafference from the efferent signal. Such mechanism, however, would require a considerable amount of neural resources and would introduce unavoidable latencies. We propose a more efficient and latency-free mechanism that does not require an efference copy but generates the movement directly together with the corresponding expected reafference. The mechanism involves a recurrent neural network that learns to generate movements from abstract movement commands, and at the same time it learns the resulting reafference from the sensory system. Afterwards, the network is able to generate both the movements together with the corresponding reafferences, and due to its intrinsic morphing capability, the network is able to flexibly interpolate and extrapolate the learned movements in synchrony with the expected reafferences. We demonstrate the modified reafference principle by computer simulations.