 |
When young children learn to write a signature, their brains generate motor commands that are different from the commands needed to do the same thing as adults. And yet, we do not need to relearn handwriting as our bones and muscles change. Our sensory-motor is equipped with a powerful machinery that modifies the previously acquired motor programs so as to compensate for changes in the body and in the surrounding environment. The activities of our research group are directed at understanding the biological and computational mechanisms underlying this remarkable ability to learn and adapt. At the same time, we wish to take advantage of the knowledge we have acquired for designing new procedures and technologies for facilitating the recovery of motor functions lost to stroke and other injuries. A neuro-robotic learning system Both the processes of learning and of adaptation take place through the plastic modification of synaptic transmission. To study how neural plasticity relates to the acquisition and adaptation of behaviors, we are developing a new experimental framework based on the interaction of a neural system with a small mobile robot. The robot and the neural system form a closed sensory-motor loop, whose observable outcome is an artificial motor behavior- for example, the robot tracking the motion of a light source. This hybrid neuro-robotic system allows us to develop and test computational models of learning and plasticity. We directly compare the artificial behaviors generated by the biological neural system with the artificial behavior generated by a mathematical model of the same system in combination with the mobile robot. Computational primitives for sensory-motor learning A number of electrophysiological studies have revealed a modular organization within the structure of the spinal cord. When a small electrical stimulation is delivered to a site in the lumbar spinal cord, a group of muscle is activated. The result of this activation is a field of forces that tend to drive the ipsilateral hindlimb toward a stable posture. Most remarkably, the simultaneous stimulation of multiple sites leads to a vectorial summation of the force fields generated by each site. These studies have suggested that the brain may generate complex motor behaviors by adding the force fields produced by multiple modules in the spinal cord. We are developing a theory of motor control based of the superposition of these force fields, which we consider as "motor primitives" in analogy with the language primitives used to generate unlimited sentences out of a finite vocabulary of words. |
|||||
| Ferdinando Mussa-Ivalldi Professor PhD, PolyTechnic of Milan |
||||||
| sandro@northwestern.edu | ||||||
| Ward 5-198 | ||||||
| (312) 503-5173 |
||||||
| Lab Website | ||||||
 |
 |
|||||
Figure: Motor adaptation to changes in arm dynamics. |
 |
 |
||||
To generate even the simplest movements of the arm, the brain must solve a complex problem of dynamics. The relation between the forces generated by the muscles of the arm and the ensuing movement is expressed by a system of complex nonlinear differential equations. A number of studies have suggested that the brain maintains an internal representation of this dynamical relation, not in the form of a mathematical expression but in the form of a transformation from desired movement to corresponding command. This type of representation has been called an "internal model". We are investigating the properties of internal models by observing how subjects interact with a robotic manipulator that applies a preprogrammed force field to their hand. This force fields constitutes a change in the dynamics that the subject's motor system must represent in order to move the hand as desired. The current experimental evidence indicate that the subjects have an accurate representation of the dependence of the force upon the state of motion of the limb. However, there is a surprising inability to represent correctly a perturbation that depends explicitly upon time. We wish to pursue this analysis to understand what are the mechanisms and the limitations of our ability to adapt to changes in limb dynamics.
How we learn to handle unstable objects Many activities of our daily lives involve handling objects which are unstable or marginally stable, such as a ladle filled with hot soup. The safe transport of the soup to one's cup is just one of the many challenges that disabled individuals Ð such as many survivors of cerebral stroke - must cope with. Handling silverware and cups, replacing light bulbs, holding a toothpaste-loaded toothbrush, hanging a shirt with a dress hanger, driving a car, all are tasks that require one to cope with some degree of instability in the manipulated object. The term ``instability'' refers to the fact that small deviations from the correct behavior of the manipulated object may lead to complete disruption of performance. For example, if because of an inappropriate movement of the wrist, the toothpaste falls off the brush, one must re-initiate the operation from scratch, which is obviously annoying. In other instances, the consequences of object instability can lead to serious damage or injury. In our laboratory we reproduce some significant features of object instability in a specially designed manipulandum. The manipulandum is programmed to generate "virtual objects" that emulate the physical presence of real objects. Looking at the performance of human subjects in transporting different virtual objects we hope to understand what mechanism are used to by the motor system to cope with mechanical instability. Designing force fields for motor rehabilitation When subjects adapt to a new mechanical environment, they learn to compensate for the disturbing forces so as to restore previously learned movements. For example, if a force disturbs our hand when we try to move it in a straight line we may produce an undesired curved movement. However, if we keep repeating the movement and the perturbation does not change, we eventually succeed in making a straight hand movement. If, at this point, the force is suddenly removed the hand will be displaced in a direction opposite to the initial curvature. This is called an "after-effect". After effects are visible manifestations of the fact that subjects have learned to anticipate the forces produced by the external perturbation. We are now investigating how after-effects may be artificially induced by specifically designed force perturbations so as to facilitate motor learning in people with motor impairments that do not involve their ability to adapt to an external force field. | ||||||
| Selected Publications: Scheidt RA, Reinkensmeyer DJ, Conditt MA, Rymer WZ, Mussa-Ivaldi FA. (2000) Persistence of motor adaptation during constrained, multi-joint, arm movements. J. Neurobiol. 84:853-862. Mussa-Ivaldi, F.A. (1999) Modular features of motor control and learning. Curr. Opin. Neurobiol. 9:713-717. Conditt MA, Mussa-Ivaldi FA. (1999) Central representation of time during motor learning. Proc. Natl. Acad. Sci. U S A. 96:11625-11630. Bizzi, E., Mussa-Ivaldi FA. (1998) The aquisition of motor behavior. Daedalus 127:217-232. Conditt MA, Gandolfo F, Mussa-Ivaldi FA. (1997) The motor system does not learn the dynamics of the arm by rote memorization of past experience. J. Neurophysiol. 78:554-560. Mussa-Ivaldi, F.A. (1997) Nonlinear force fields: a distributed system of control primitives for representing and learning movements. Proceedings of the 1997 IEEE International Symposium on Computational Intelligence in Robotics and Automation. p. 84-90. Conditt, M.A., Gandolfo, F. and Mussa-Ivaldi, F.A. (1997) The motor system does not learn the dynamics of the arm by rote memorization of past experience. J. Neurophysiol. 78: 554-560. Mussa-Ivaldi, F.A. (1995) Geometrical Principles in Motor Control. In: M.A. Arbib (Ed.) Handbook of Brain Theory. MIT Press, Cambridge, MA. Mussa-Ivaldi, F.A., Giszter, F.A. and Bizzi, E. (1994) Linear combination of primitives in vertebrate motor control. Proc. Nat. Acad. Sci. USA 91: 7534-7538. Shadmehr, R. and Mussa-Ivaldi, F.A. (1994) Adaptive representation of dynamics during learning of a motor task. J. Neurosci.14: 3208-3224.
|
||||||
| home | about | research | faculty | students | postDocs | research associates | staff | news | semiars | employment | visit | links | search | sitemap |