Projects/ Research areas
What is sensorimotor control?
Sensorimotor control [lat. sensus = sense; the ability to percept; motor = a mover]. Human motor
control studies all voluntary movements initiated and regulated by the central nervous system.
Sensorimotor control concerns relationships between sensation and movement or , more broadly,
between perception and action. The interplay of sensory and motor processes provides the basis of
observable human behavior. Wilhelm Wundt (1858), the founder of Physiological Psychology,
had realized that an understanding of complex behavioral patterns will be
impossible without understanding the simple elements. Consequently, in trying to comprehend
general human behavior, motor control and perception are integral parts of the curriculum of a
student in psychology.
Focus 1: Modelling and Learning the Inverse Dynamics
and Kinematics of the Arm
The central question of our work is: What parameters are controlled by the nervous system to
achieve limb coordination? We are especially concerned with the motor control of arm
movements. In the past years physically correct models of the human arm have been developed in
computer simulations, which can be controlled by artificial "neural" networks.
In addition to the simulation work, experiments examine the kinematics, dynamics, and muscular
patterns of goal-directed arm movements. For our experiments subjects perform single-joint arm
movements by manipulating a lever. The physical make-up of the arm-lever system can be
perturbated by manipulating inertia, damping, and stiffness electronically.
 |
Partial view of the experimental setup:
Subjects perform goal-directed elbow movements. (Stimulus setup not shown here). These movements
are perturbed or manipulated by generating additional torque components via the motor underneath the lever (grey cylinder).
|
 |
Partial view of the experimental setup:
We record the electromyographic activity of elbow flexors and exentsors next to measuring the dynamics and kinematics.
|
Supported by the Deutsche Forschungsgemeinschaft (German Science Foundation)
Focus 2: Development of motor control in ontogenesis
Few studies engage in the developmental aspect of motor control in children at the age of four to eleven
years. We investigate how and when children learn to control movements by way of an inverse model of
the arm. To this end, we study the process of adaptating to external forces when performing forearm
movements, as well as the dependence on continuous visual feedback. Slow adaptation or worsening of motor control
in the absence of continuous visual feedback speaks in favour of unprecise inverse models. Furthermore,
if adaptation reflects an update of internal motor models, one would expect that learning should transfer
to other kinds of movements as compared to those of the learning phase.
Sensorimotor Development in Biological and Artificial Systems
(Cooperation with the Laboratory of Integrated Advanced Robotics, University of Genova, Italy)
In this project we apply our knowledge about the human motor system to the world of artificial robots. We advocate
that the framework of biological development is suitable for the construction of artificial systems.
The LIRA lab in Genova (Prof. G. Sandini) has pursued this issue for some time and constructed a
robot system that can reach for objects in the environment. In contrast to most neural networks, the system controller
does not have explicit knowledge of the arm kinematics or dynamics. The arm system is initially controlled by infant-like
"newborn" reflexes.
 |
|
 |
| Video image of the arm robot reaching for a target. |
|
Example of a newborn reflex. Here two infants exhibit the so-called "fencer reflex"
[asymetric-tonic-neck-reflex]. Turning the head elicits arm extension and contralateral arm flexion. |
To visit the Italian group, click below:
 University of Genova - Dipartimento di Informatica, Sistemistica e Telematica
http://www.lira.dist.unige.it
Supported by the Deutsche Forschungsgemeinschaft (German Science Foundation)
Supported by German-Italian Research Exchange Program VIGONI
Focus 3: Sensorimotor Control of Arm Movements
in Patients with Brain Injury
(Cooperation with the Neurology Departments of the Universities Essen, Düsseldorf,
and Tübingen)
Here we investigate how the control of the upper extremities changes as a result
of specific lesions in the neocortex and the basal ganglia. We ask the patients to perform
simple arm movements and record these movements and the underlying electromyographic signals
from several arm muscles. Based on the specific deficits seen in these patients, we attempt
to make inferences about the function of the lesioned brain structures for the control of
voluntary movement. This is a difficult endeavour, as there clearly is not always a one-to-one
mapping between lesion and the observed behavioural deficit.
|
The figure shows the time course of the movement of three subjects during
forearm target movement. Note the slowing of the movement in the Parkinsonian patients resp.
the hypermetrie (overshoot) in the cerebellar patients' movements.
|
Focus 4: Modelling of human
motor control
In this project we try to simulate control of arm movement by way of a mass-spring-model
and an Inverse model. The system's biomechanics are represented in the
mass-spring-model, the Inverse model represents the neuronal controller.
The crucial problem lies in the fact that there is no unique solution to an inverse computation:
the same movement can be produced through an infinite number of combinations of parameters of
stiffness, damping and equilibrium position.
Simulation of neuronal movement control requires further assumptions and simplifications.
Here we show that movements can be controlled through changes in equilibrium position only,
if damping and stiffness coefficients are held constant.
The numerical values used in the simulation are extracted from experimental data. As movement
plan we used mean kinemaics, that is angular position, velocity and acceleration from baseline
movements. To get an idea of the order of magnitude of stiffness and damping parameters, we used a
specific algorithm for parameter-identification. For the course of the
simulation see block diagram.
If the Inverse Model is correct, i.e. the same parameters are set in the simulation and the real model of the arm, then the resulting movement should equal the movement plan. If there are external perturbations (for example force pulses), though, which are not included in the inverse model, then a deviation between simulated and real data will result. Model predictions can be tested with experimental data where force pulses were applied. (plots:
unperturbed movements /
perturbed movements)
Parameter-Identification: Infering
from measured kinematics (angular position, velocity and acceleration) and external
forces to the parameters set by the organism, constitutes an ill posed problem, because there ia an infinite number of possible solutions. For each timed sample, there are three unknown variables in the differential equation describing the physical system. We chose a least square algorithm to solve the problem. We suppose that parameters vary only slowly over time and thus can be considered constant in small time windows. Then there is an overdetermined system of equations for the parameters (for each timed sample one considers information from previous and following states of the system. From this system of equation, the parameters can be determined by (pseudo-) matrix inversion (Plots for Comparison
between simulation and original movement; literature:
Konczak,Brommann & Kalveram 1999 )
Supported by the Deutsche Forschungsgemeinschaft (German Science Foundation)
http://www.bio.psy.ruhr-uni-bochum.de/spp/
|
