070920 Mergner

Prof. Dr. Thomas Mergner

Neurozentrum University Medical Center Freiburg

First, I report principles of multisensory control of human upright stance which we derived from psychophysics of self-motion perception. The principles allowed us to develop a simple sensory feedback model that differs from similar attempts in the literature in that it is based on feedback of external disturbance estimates (external physical stimuli) instead of sensor signals. The estimates are derived from intersensory interactions and thresholding. They do not involve an efference copy-like signal as it is used in the optimal estimator approach. The model describes experimental stance control data in various behavioural situations. Different types of disturbance may be superimposed on each other and on voluntary movement. The model design allows for human-like low loop gain and long delays. Embodied into a biped humanoid robot ("PostuRob"), the model demonstrates its validity in the ‘real world’ (sensor noise, drifts, etc.). Through this embodiment we want to develop new hypotheses on posture control deficits in neurological patients, therapy effects, and rehabilitation aims. In a second part of the talk, I show that the approach holds not only for steady state stimulus responses, but upon inclusion of few model extensions also for transient responses, adaptation, etc. When omitting in the model and robot the inertial sensors that represent the vestibular system, we could use them to mimic experimental data from vestibular loss patients and to forward a hypothesis on how they are able to cope, eyes closed, even with support surface tilt stimuli despite the impoverished sensory inputs (essentially only joint angle and force/joint torque inputs).