Single trial learning of external dynamics: What can the brain teach us about learning mechanisms?

Humans have exceptional abilities to learn new skills, manipulate tools and objects, and interact with our environment. In order to be successful at these tasks, our brain has developed learning mechanisms to deal with and compensate for the constantly changing dynamics of the world. If this mechanism or mechanisms can be understood from a computational point of view, then they can also be used to drive the adaptability and learning of robots. In this paper, we will present a new technique for examining changes in the feedforward motor command due to adaptation. This technique can then be utilized for examining motor adaptation in humans and determining a computational algorithm which explains motor learning. © 2007.

Learning the dynamics of the external world: Brain inspired learning for robotic applications

Humans have exceptional abilities to learn new skills, manipulate tools and objects, and interact with our environment. In order to be successful at these tasks, our brain has become exceptionally well adapted to learning to deal not only with the complex dynamics of our own limbs but also with novel dynamics in the external world. While learning of these dynamics includes learning the complex time-varying forces at the end of limbs through the updating of internal models, it must also include learning the appropriate mechanical impedance in order to stabilize both the limb and any objects contacted in the environment. This article reviews the field of human learning by examining recent experimental evidence about adaptation to novel unstable dynamics and explores how this knowledge about the brain and neuro-muscular system can expand the learning capabilities of robotics and prosthetics. © 2006.

Flexible representations of dynamics are used in object manipulation.

To manipulate an object skillfully, the brain must learn its dynamics, specifying the mapping between applied force and motion. A fundamental issue in sensorimotor control is whether such dynamics are represented in an extrinsic frame of reference tied to the object or an intrinsic frame of reference linked to the arm. Although previous studies have suggested that objects are represented in arm-centered coordinates [1-6], all of these studies have used objects with unusual and complex dynamics. Thus, it is not known how objects with natural dynamics are represented. Here we show that objects with simple (or familiar) dynamics and those with complex (or unfamiliar) dynamics are represented in object- and arm-centered coordinates, respectively. We also show that objects with simple dynamics are represented with an intermediate coordinate frame when vision of the object is removed. These results indicate that object dynamics can be flexibly represented in different coordinate frames by the brain. We suggest that with experience, the representation of the dynamics of a manipulated object may shift from a coordinate frame tied to the arm toward one that is linked to the object. The additional complexity required to represent dynamics in object-centered coordinates would be economical for familiar objects because such a representation allows object use regardless of the orientation of the object in hand.

Functional significance of stiffness in adaptation of multijoint arm movements to stable and unstable dynamics.

This study compared the mechanisms of adaptation to stable and unstable dynamics from the perspective of changes in joint mechanics. Subjects were instructed to make point to point movements in force fields generated by a robotic manipulandum which interacted with the arm in either a stable or an unstable manner. After subjects adjusted to the initial disturbing effects of the force fields they were able to produce normal straight movements to the target. In the case of the stable interaction, subjects modified the joint torques in order to appropriately compensate for the force field. No change in joint torque or endpoint force was required or observed in the case of the unstable interaction. After adaptation, the endpoint stiffness of the arm was measured by applying displacements to the hand in eight different directions midway through the movements. This was compared to the stiffness measured similarly during movements in a null force field. After adaptation, the endpoint stiffness under both the stable and unstable dynamics was modified relative to the null field. Adaptation to unstable dynamics was achieved by selective modification of endpoint stiffness in the direction of the instability. To investigate whether the change in endpoint stiffness could be accounted for by change in joint torque or endpoint force, we estimated the change in stiffness on each trial based on the change in joint torque relative to the null field. For stable dynamics the change in endpoint stiffness was accurately predicted. However, for unstable dynamics the change in endpoint stiffness could not be reproduced. In fact, the predicted endpoint stiffness was similar to that in the null force field. Thus, the change in endpoint stiffness seen after adaptation to stable dynamics was directly related to changes in net joint torque necessary to compensate for the dynamics in contrast to adaptation to unstable dynamics, where a selective change in endpoint stiffness occurred without any modification of net joint torque.

Different mechanisms involved in adaptation to stable and unstable dynamics.

Recently, we demonstrated that humans can learn to make accurate movements in an unstable environment by controlling magnitude, shape, and orientation of the endpoint impedance. Although previous studies of human motor learning suggest that the brain acquires an inverse dynamics model of the novel environment, it is not known whether this control mechanism is operative in unstable environments. We compared learning of multijoint arm movements in a "velocity-dependent force field" (VF), which interacted with the arm in a stable manner, and learning in a "divergent force field" (DF), where the interaction was unstable. The characteristics of error evolution were markedly different in the 2 fields. The direction of trajectory error in the DF alternated to the left and right during the early stage of learning; that is, signed error was inconsistent from movement to movement and could not have guided learning of an inverse dynamics model. This contrasted sharply with trajectory error in the VF, which was initially biased and decayed in a manner that was consistent with rapid feedback error learning. EMG recorded before and after learning in the DF and VF are also consistent with different learning and control mechanisms for adapting to stable and unstable dynamics, that is, inverse dynamics model formation and impedance control. We also investigated adaptation to a rotated DF to examine the interplay between inverse dynamics model formation and impedance control. Our results suggest that an inverse dynamics model can function in parallel with an impedance controller to compensate for consistent perturbing force in unstable environments.

Central representation of dynamics when manipulating handheld objects.

To explore the neural mechanisms related to representation of the manipulation dynamics of objects, we performed whole-brain fMRI while subjects balanced an object in stable and highly unstable states and while they balanced a rigid object and a flexible object in the same unstable state, in all cases without vision. In this way, we varied the extent to which an internal model of the manipulation dynamics was required in the moment-to-moment control of the object's orientation. We hypothesized that activity in primary motor cortex would reflect the amount of muscle activation under each condition. In contrast, we hypothesized that cerebellar activity would be more strongly related to the stability and complexity of the manipulation dynamics because the cerebellum has been implicated in internal model-based control. As hypothesized, the dynamics-related activation of the cerebellum was quite different from that of the primary motor cortex. Changes in cerebellar activity were much greater than would have been predicted from differences in muscle activation when the stability and complexity of the manipulation dynamics were contrasted. On the other hand, the activity of the primary motor cortex more closely resembled the mean motor output necessary to execute the task. We also discovered a small region near the anterior edge of the ipsilateral (right) inferior parietal lobule where activity was modulated with the complexity of the manipulation dynamics. We suggest that this is related to imagining the location and motion of an object with complex manipulation dynamics.