Influence of haptic guidance in learning a novel visuomotor task

Abstract

In (re)learning of movements, haptic guidance can be used to direct the needed adaptations in motor control. Haptic guidance influences the main driving factors of motor adaptation, execution error, and control effort in different ways. Human-control effort is dissipated in the interactions that occur during haptic guidance. Minimizing the control effort would reduce the interaction forces and result in adaptation. However, guidance also decreases the magnitude of the execution errors, which could inhibit motor adaptation. The aim of this study was to assess how different types of haptic guidance affect kinematic adaptation in a novel visuomotor task. Five groups of subjects adapted to a reaching task in which the visual representation of the hand was rotated 30°. Each group was guided by a different force field. The force fields differed in magnitude and direction in order to discern the adaptation based on execution errors and control effort. The results demonstrated that the execution error did indeed play a key role in adaptation. The more the guiding forces restricted the occurrence of execution errors, the smaller the amount and rate of adaptation. However, the force field that enlarged the execution errors did not result in an increased rate of adaptation. The presence of a small amount of adaptation in the groups who did not experience execution errors during training suggested that adaptation could be driven on a much slower rate and on the basis of minimization of control effort as was evidenced by a gradual decrease of the interaction forces during training. Remarkably, also in the group in which the subjects were passive and completely guided, a small but significant adaptation occurred. The conclusion is that both minimization of execution errors and control effort drives kinematic adaptation in a novel visuomotor task, but the latter at a much slower rate.

Introduction

Haptic guidance of movements can be used to demonstrate to a subject how fast and in which direction a movement should be performed. As such, haptic guidance is used for learning new skills in sports, but also for relearning motor control after a stroke (Kahn et al., 2006). In the latter case, the guidance was traditionally applied manually by a therapist. However, in the last decade, different robotic devices for neurorehabilitation (Colombo et al., 2000, Ferraro et al., 2003, Hogan and Krebs, 2004, Lum et al., 2002) have been developed that can provide continuous guidance during the recommended highly repetitive practicing (Kwakkel et al., 2004, Teasell et al., 2005).

With these therapy robots, different types of haptic guidance have been implemented. Soft guidance moves a limb through a pre-specified trajectory where deviations from this trajectory result in forces toward this trajectory (Aisen et al., 1997). For hard guidance, haptic tunnels are rendered. A subject can move within this tunnel (Kahn et al., 2006) but not outside the stiff walls of the tunnel. In passive guidance, the robot is position-controlled and enforces a pre-specified trajectory. Consequently, the subject does not need to contribute to the movement and can be fully passive (Hesse et al., 2003, Lynch et al., 2005). Finally, haptic force fields have also been used to increase, instead of reduce, errors in the execution of movement (Emken and Reinkensmeyer, 2004, Patton et al., 2006, Wei et al., 2005).

Our interest is to increase the understanding of the interactions between haptic guidance and the learning of a novel motor task. In the gross of computational motor control theories, the underlying principle is the minimization of both task-execution errors and control effort, (Shadmehr and Krakauer, 2008, Todorov, 2004, van der Kooij et al., 1999), which can be defined as the required muscular energy to complete a movement. These latter two factors are also considered to be the driving factors of motor learning in different computational theories of motor learning (Franklin et al., 2008, Stroeve, 1996). Experimental studies show that minimization of both control effort and execution error indeed characterizes learning dynamics (Emken et al., 2007a, Scheidt et al., 2000). These studies made use of haptic devices, similar to those being used in neurorehabilitation, to study how subjects adapt when exposed to force fields applied to moving limbs. In the study of Scheidt and colleagues (2000), subjects first learned to reach in a viscous force field. Subsequently, the viscous force field was removed and kinematic after effects were prevented from occurring by using a rendered haptic tunnel. They showed that subjects made reaching movements while simultaneously exerting perpendicular forces to the haptic channel. These forces were similar to the forces required to compensate for the viscous force field.

Despite the absence of kinematic execution errors, subjects disadapted by decaying the forces exerted on the channel over the different movements. Still, the disadaptation occurred at a much slower rate than when kinematic errors were allowed to occur. Further evidence for a contribution of muscular effort in adaptation was recently provided by Emken and colleagues (2007b). They examined the adaptation to an externally applied force field during the swing phase of walking and showed that a model describing the temporal evolution of error (Scheidt et al., 2001, Thoroughman and Shadmehr, 2000) could be derived from minimization of a cost function, that is a weighted sum of the execution error and control effort.

Haptic guidance influences the driving factors of motor control and learning in different ways. Guidance directly influences the execution errors, either reducing or enlarging these errors. Though in a more indirect way, guidance also affects the control effort. Generally, guidance is only applied when the movement deviates from the optimal trajectory. The exerted forces against the walls of the wall are considered to be a dissipation of effort. Minimizing these interaction forces would result in the most efficient movement and therefore in the required adaptation. It is not yet known how these manipulations of the driving factors interfere with the process of motor adaptation. The aim of this study was to assess how external guidance influences motor adaptation.

In previously discussed studies, motor adaptation was studied by exposing subjects to a force perturbation that simulates a viscous load. This kind of perturbation is unsuitable for studying the net effect of guiding force fields as the guiding forces would interfere with these disturbing forces. Therefore, we applied different types of haptic guidance while subjects adapted to a kinematic distortion, that is, a visuomotor task. In this task, subjects made centre-out reaching movements while the visual feedback of hand-movement direction was rotated 30° counter clockwise by making use of a virtual reality setup (Caithness et al., 2004, Ghez et al., 2000, Krakauer et al., 1999, Sainburg and Wang, 2002, Tong et al., 2002, Wang and Sainburg, 2005). To discern the adaptation based on execution errors and control effort, different types of haptic guidance that are commonly used in therapy robots were applied. All but one of these force fields applied forces only in the direction perpendicular to the target direction, which necessitates the subjects to move in the target direction themselves. Subjects in the error-enhanced group (EE) received hand forces that were proportional and in the same direction as the execution errors, defined as the deviations from the straight path toward the target. These forces effectively enlarged the execution errors. In the soft (SG) and hard guidance (HG) groups, error-correcting forces were applied to the hand that were proportional but opposite to the execution errors. In the soft guidance group, the low stiffness of the force field still allowed considerable execution errors. However, in the hard guidance group, the high stiffness formed a haptic tunnel, denying all but very small deviations (<1.5 mm) from the optimal trajectory. This tunnel was similar to the tunnel used by Scheidt and colleagues (2000). In the passive group (P), the subjects were moved along the optimal trajectory by the robot and were instructed not to intervene. In this case, execution errors were zero and the control effort did not influence task performance. In the control group (A), no additional guidance was applied.

Previous studies on adaptation to force perturbations showed that adaptation based on execution errors occurred at a faster rate than adaptation based on minimization of effort. Therefore, we hypothesize that when execution errors are increased (EE) or reduced (SG) by haptic guidance, the rate of adaptation will be faster or slower respectively, and in both cases adaptation will be complete. In case execution errors are prevented (HG) but subjects have to actively move their hand toward the target, we hypothesize that adaptation still occurs but at a much slower rate than in the A, EH, and SG groups. In this case, adaptation is driven by minimization of control effort solely. For the passive group, we hypothesize that adaptation is absent since no execution errors occur and control effort is not related to task instruction and performance.