Error generalization as a function of velocity and duration: human reaching movements

Learning context shapes bimanual control strategy and generalization of novel dynamics

Bimanual movements are fundamental components of everyday actions, yet the underlying mechanisms coordinating adaptation of the two hands remain unclear. Although previous studies highlighted the contextual effect of kinematics of both arms on internal model formation, we do not know how the sensorimotor control system associates the learned memory with the experienced states in bimanual movements. More specifically, can, and if so, how, does the sensorimotor control system combine multiple states from different effectors to create and adapt a motor memory? Here, we tested motor memory formation in two groups with a novel paradigm requiring the encoding of the kinematics of the right hand to produce the appropriate predictive force on the left hand. While one group was provided with training movements in which this association was evident, the other group was trained on conditions in which this association was ambiguous. After adaptation, we tested the encoding of the learned motor memory by measuring the generalization to new movement combinations. While both groups adapted to the novel dynamics, the evident group showed a weighted encoding of the learned motor memory based on movements of the other (right) hand, whereas the ambiguous group exhibited mainly same (left) hand encoding in bimanual trials. Despite these differences, both groups demonstrated partial generalization to unimanual movements of the left hand. Our results show that motor memories can be encoded depending on the motion of other limbs, but that the training conditions strongly shape the encoding of the motor memory formation and determine the generalization to novel contexts.

On the encoding capacity of human motor adaptation

Primitive-based models of motor learning suggest that adaptation occurs by tuning the responses of motor primitives. Based on this idea, we consider motor learning as an information encoding procedure, that is, a procedure of encoding a motor skill into primitives. The capacity of encoding is determined by the number of recruited primitives, which depends on how many primitives are "visited" by the movement, and this leads to a rather counterintuitive prediction that faster movement, where a larger number of motor primitives are involved, allows learning more complicated motor skills. Here, we provide a set of experimental results that support this hypothesis. First, we show that learning occurs only with movement, that is, only with nonzero encoding capacity. When participants were asked to counteract a rotating force applied to a robotic handle, they were unable to do so when maintaining a static posture but were able to adapt when making small circular movements. Our second experiment further investigated how adaptation is affected by movement speed. When adapting to a simple (low-information-content) force field, fast (high-capacity) movement did not have an advantage over slow (low-capacity) movement. However, for a complex (high-information-content) force field, the fast movement showed a significant advantage over slow movement. Our final experiment confirmed that the observed benefit of high-speed movement is only weakly affected by mechanical factors. Taken together, our results suggest that the encoding capacity is a genuine limiting factor of human motor adaptation.NEW & NOTEWORTHY We propose a novel concept called "encoding capacity" of motor adaptation, which describes an inherent limiting-factor of our brain's ability to learn new motor skills, just like any other storage system. By reinterpreting the existing primitive-based models of motor learning, we hypothesize that the encoding capacity is determined by the size of the movement, and present a set of experimental evidence suggesting that such limiting effect of encoding capacity does exist in human motor adaptation.

An adaptive spinal-like controller: tunable biomimetic behavior for a robotic limb

Background

Spinal-like regulators have recently been shown to support complex behavioral patterns during volitional goal-oriented reaching paradigms. We use an interpretation of the adaptive spinal-like controller as inspiration for the development of a controller for a robotic limb. It will be demonstrated that a simulated robot arm with linear actuators can achieve biological-like limb movements. In addition, it will be shown that programmability in the regulator enables independent spatial and temporal changes to be defined for movement tasks, downstream of central commands using sensory stimuli. The adaptive spinal-like controller is the first to demonstrate such behavior for complex motor behaviors in multi-joint limb movements.

Methods

The controller is evaluated using a simulated robotic apparatus and three goal-oriented reaching paradigms: 1) shaping of trajectory profiles during reaching; 2) sensitivity of trajectories to sudden perturbations; 3) reaching to a moving target. The experiments were designed to highlight complex motor tasks that are omitted in earlier studies, and important for the development of improved artificial limb control.

Results

In all three cases the controller was able to reach the targets without a priori planning of end-point or segmental motor trajectories. Instead, trajectory spatio-temporal dynamics evolve from properties of the controller architecture using the spatial error (vector distance to goal). Results show that curvature amplitude in hand trajectory paths are reduced by as much as 98% using simple gain scaling techniques, while adaptive network behavior allows the regulator to successfully adapt to perturbations and track a moving target. An important observation for this study is that all motions resemble human-like movements with non-linear muscles and complex joint mechanics.

Conclusions

The controller shows that it can adapt to various behavioral contexts which are not included in previous biomimetic studies. The research supplements an earlier study by examining the tunability of the spinal-like controller for complex reaching tasks. This work is a step toward building more robust controllers for powered artificial limbs.

A Simplified Spinal-Like Controller Facilitates Muscle Synergies and Robust Reaching Motions

We develop an adaptive controller for multi-joint, multi-muscle arm movements based on simplified spinal-like circuits found in the periphery, muscle synergies, and interpretations of gain-field projections from reach related neurons in the Superior Colliculus. The resulting innovation provides a highly robust sensory based controller that can be adapted to systems which require multi-muscle co-ordination. It provides human-like responses during perturbations elicited either internally or by the environment and for simple point-to-point reaching. We simulate limb motion and EMGs in Simulink using Virtual Muscle models and a variety of paradigms, including motion with external perturbations, and varying levels of antagonist muscle co-contractions. The results show that the system can exhibit smooth coordinated motions, without explicit kinematic or dynamic planning even in the presence of perturbations. In addition, we show by varying the level of muscle co-contractions from 0% to 40%, that the effects of external perturbations on joint trajectories can be reduced by up to 42%. The improved controller design is novel providing robust behavior during dynamic events and an automatic adaptive response from sensory-integration.

Using dual tasks to test immediate transfer of training between naturalistic movements: a proof-of-principle study

Theories of motor learning predict that training a movement reduces the amount of attention needed for its performance (i.e., more automatic). If training one movement transfers, then the amount of attention needed for performing a second movement should also be reduced, as measured under dual task conditions. The authors' purpose was to test whether dual task paradigms are feasible for detecting transfer of training between two naturalistic movements. Immediately following motor training, subjects improved performance of a second untrained movement under single and dual task conditions. Subjects with no training did not. Improved performance in the untrained movement was likely due to transfer, and suggests that dual tasks may be feasible for detecting transfer between naturalistic actions.

Influence of planar manipulandum to the hand trajectory during point to point movement

We present the analysis of the planar manipulandum effects to the trajectory of point to point movements in horizontal plane. This analysis is of significance for the control of a haptic robot that can be used for the rehabilitation of hemiplegic patients. The effects were assessed by comparing data collected in experiments with healthy subjects when performing simple movements that are used in the therapy of stroke patients. We found significant differences between the preferred trajectories and the deviations from the preferred trajectories (p<0.01) when moving with and without the manipulandum. This result suggests that for the design of the controller of a robot assistant inertial properties of the robot mechanism must be considered even in the case that it is used only for the assessment (passive) or within the bio-feedback.

Generalization and multirate models of motor adaptation

When subjects adapt their reaching movements in the setting of a systematic force or visual perturbation, generalization of adaptation can be assessed psychophysically in two ways: by testing untrained locations in the work space at the end of adaptation (slow postadaptation generalization) or by determining the influence of an error on the next trial during adaptation (fast trial-by-trial generalization). These two measures of generalization have been widely used in psychophysical studies, but the reason that they might differ has not been addressed explicitly. Our goal was to develop a computational framework for determining when a two-state model is justified by the data and to explore the implications of these two types of generalization for neural representations of movements. We first investigated, for single-target learning, how well standard statistical model selection procedures can discriminate two-process models from single-process models when learning and retention coefficients were systematically varied. We then built a two-state model for multitarget learning and showed that if an adaptation process is indeed two-rate, then the postadaptation generalization approach primarily probes the slow process, whereas the trial-by-trial generalization approach is most informative about the fast process. The fast process, due to its strong sensitivity to trial error, contributes predominantly to trial-by-trial generalization, whereas the strong retention of the slow system contributes predominantly to postadaptation generalization. Thus, when adaptation can be shown to be two-rate, the two measures of generalization may probe different brain representations of movement direction.

Linear hypergeneralization of learned dynamics across movement speeds reveals anisotropic, gain-encoding primitives for motor adaptation

The ability to generalize learned motor actions to new contexts is a key feature of the motor system. For example, the ability to ride a bicycle or swing a racket is often first developed at lower speeds and later applied to faster velocities. A number of previous studies have examined the generalization of motor adaptation across movement directions and found that the learned adaptation decays in a pattern consistent with the existence of motor primitives that display narrow Gaussian tuning. However, few studies have examined the generalization of motor adaptation across movement speeds. Following adaptation to linear velocity-dependent dynamics during point-to-point reaching arm movements at one speed, we tested the ability of subjects to transfer this adaptation to short-duration higher-speed movements aimed at the same target. We found near-perfect linear extrapolation of the trained adaptation with respect to both the magnitude and the time course of the velocity profiles associated with the high-speed movements: a 69% increase in movement speed corresponded to a 74% extrapolation of the trained adaptation. The close match between the increase in movement speed and the corresponding increase in adaptation beyond what was trained indicates linear hypergeneralization. Computational modeling shows that this pattern of linear hypergeneralization across movement speeds is not compatible with previous models of adaptation in which motor primitives display isotropic Gaussian tuning of motor output around their preferred velocities. Instead, we show that this generalization pattern indicates that the primitives involved in the adaptation to viscous dynamics display anisotropic tuning in velocity space and encode the gain between motor output and motion state rather than motor output itself.