Lecture Notes in Computer Science
Reflex Contributions to the Directional Tuning
Download 12.42 Mb. Pdf ko'rish
|
- Bu sahifa navigatsiya:
- 2 Methods
- 2.2 Force Fields
- 2.3 Protocol
- 3 Results
- 3.1 Endpoint Stiffness
- 3.2 Electromyographic Activity
Reflex Contributions to the Directional Tuning
of Arm Stiffness Gary Liaw 1 , David W. Franklin 1,2,3 , Etienne Burdet 4 , Abdelhamid Kadi-allah 4 ,
and Mitsuo Kawato 3 1 National Institute of Information and Communications Technology, 2-2-2 Hikaridai, Keihanna Science City, Kyoto, 619-0288, Japan 2 Department of Engineering, University of Cambridge, Cambridge, United Kingdom 3 ATR Computational Neuroscience Laboratories, Keihanna Science City, Kyoto, Japan 4 Department of Bioengineering, Imperial College London, London, United Kingdom Abstract. It has been shown that during arm movement, humans selectively change the endpoint stiffness of their arm to compensate for the instability in an unstable environment. When the direction of the instability is rotated with respect to the direction of movement, it was found that humans modify the antisymmetric component of their endpoint stiffness. The antisymmetric component of stiffness arises due to reflex responses suggesting that the subjects may have tuned their reflex responses as part of the feedforward adaptive control. The goal of this study was to examine whether the CNS modulates the gain of the reflex response for selective tuning of endpoint impedance. Subjects performed reaching movements in three unstable force fields produced by a robotic manipulandum, each field differing only in the rotational component. After subjects had learned to compensate for the field, allowing them to make unperturbed movements to the target, the endpoint stiffness of the arm was estimated in the middle of the movements. At the same time electromyographic activity (EMG) of six arm muscles was recorded. Analysis of the EMG revealed differences across force fields in the reflex gain of these muscles consistent with stiffness changes. This study suggests that the CNS modulates the reflex gain as part of the adaptive feedforward command in which the endpoint impedance is selectively tuned to overcome environmental instability. Keywords: Mechanical impedance, limb stiffness, internal model, stretch reflex, impedance control, co-contraction, electromyography. 1 Introduction In everyday life, we perform activities in our environment that require us to interact with different objects such as tools. These interactions with tools are often inherently unstable [1]. For example, when using a screwdriver, the direction of the force applied needs to be parallel to the screwdriver in order to maintain contact with the screw. However this is further complicated as the force applied by the user is susceptible to fluctuations due to inherent signal dependent noise [2]. These variations
914 G. Liaw et al. in the direction of the applied force could produce movements that may cause the screwdriver to slip, and lose contact with the screw. In order to successfully perform this and other unstable tasks, the neuromuscular system must overcome such instability. Viscoeleastic properties of muscle play an important role in motor control, as it reacts without delay to disturbances caused by instabilities. The greater the viscoelastic impedance of the arm, the better it can resist perturbations that affect the intended position or trajectory. The ability to control viscoelastic impedance is therefore important, especially in unstable environments [3] or unpredictable circumstances [4]. Hogan first suggested that impedance can be selectively controlled by the central nervous system (CNS) [5] and evidence for this selective control was later demonstrated by others [3,6,7]. In the latter studies, point-to-point arm reaching movements were used as a platform to investigate the level of sophistication of impedance control by the CNS. It was demonstrated that arm stiffness could be modulated by the CNS for magnitude [6] as well as direction [7] of the environmental instability. It was suggested that such modulation was facilitated by feedforward changes in muscle activation, specifically by modifying the activations in three muscle groups: the shoulder muscles, the elbow muscles and the biarticular muscles. Endpoint stiffness can be decomposed into two components: a symmetric component and an antisymmetric component [8]. The symmetric component can be produced by a combination of passive, intrinsic, or symmetric reflexive stiffness, while the antisymmetric component is thought to be mainly due to heteronymous reflexive responses [5]. This means that a large antisymmetric component in the endpoint stiffness can be interpreted as evidence of contribution from reflexes. Our previous study [7] found that adaptation to environmental instability in different directions resulted in changes to the antisymmetric component of endpoint stiffness. The present study aims at solidifying those results by varying only the cross terms of the environmental instability. We will investigate whether the CNS modulates the reflex contributions to stiffness by modifying the reflex gain as part of the impedance controller.
Four healthy, right-handed subjects (three male, one female) participated in the study. The experiment was approved by the institutional ethics committee and subjects gave informed consent. 2.1 Apparatus Subjects were seated in a chair with their shoulder held against the back of the chair by seatbelts to restrict trunk movement. The height of the chair was adjusted such that the workspace was at the subject’s shoulder level. A custom-molded thermoplastic cuff was used to restrict the subject’s wrist motion and a horizontal beam was secured to the subject’s forearm for support. The subject’s right arm, along with the cuff and the beam were coupled to the parallel-link direct drive air-magnet floating manipulandum (PFM) (Fig 1). Details of the PFM and setup can be found elsewhere [9]. Reflex Contributions to the Directional Tuning of Arm Stiffness 915 Subjects performed point-to-point reaching movements in the horizontal plane from a start target to an end target. The start and end targets were marked with a circle of diameters 2.5cm and 3cm, respectively, and located 31cm and 56cm, respectively, from and directly in front of the shoulder joint. The subject’s arm was restricted to keep the movement to two degrees of freedom. Subjects were asked to perform the movements in 600ms ± 100ms. A computer monitor provided feedback to the subjects on whether the previous movement was successful or not. During force field trials, safety boundaries were set up 5cm from the center on either side, beyond which the force field would revert to a null field. A successful movement was one where the subject completed the movement by reaching the target within the desired duration range, and did not exceed the safety boundaries. An opaque table above the workspace blocked the subject's view of the manipulandum and the subject's arm. The position of the hand and the start and end target circles were projected onto the surface of the table to provide subjects with visual feedback. PFM
target start
y x [0,0] Fig. 1. Experimental Setup. The subject’s right arm was attached to the handle of the PFM with a cuff, restricting wrist motion. Reaching movement were performed from a start position [0, 31]cm to a target position [0, 56]cm relative to the shoulder joint. The PFM either applied a computer-controlled force field or controlled displacements for stiffness estimation. 2.2 Force Fields Subjects performed point-to-point reaching movements in three different unstable divergent force fields as well as a null field. The three force fields exerted a force away from the center line (y-axis) of the workspace that is proportional to the size of the deviation from the center line; the subjects experienced no force on the hand if there is no deviation. The three force fields differed in direction and magnitude. As a baseline for adaptation, a simple divergent force field (DF), oriented perpendicular to the direction of movement, was simulated. The equation was: 200 0
0 x y F x F y ⎡ ⎤ ⎡
⎤ ⎡ ⎤ = ⋅ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ ⎦ ⎣ ⎦
⎣ ⎦
(1) 916 G. Liaw et al. where F
are the forces applied by the PFM in the x and y directions, respectively. The other two divergent force fields simply varied in terms of the cross term linking an error in x with the force produced in y. Each of these rotated the forces either in the clockwise direction (CW DF) or counter clockwise direction (CCW DF). The equations of the force fields were: CW DF ; CCW DF
200 0 200 0 : :
600 0 600 0 x x y y F F x x F F y y ⎡ ⎤
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
= ⋅ = ⋅ ⎢ ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
− ⎣ ⎦ ⎣ ⎦ ⎣ ⎦ ⎣ ⎦
⎣ ⎦ ⎣ ⎦
(2)
In order to compensate for the imposed force fields, it has been shown that subjects increase the limb stiffness to compensate for the instability of the environment [3,6]. In order to increase one cross term (eg. K yx ) without modifying the other one (eg. K xy ), it has been suggested that subjects change the reflex gain [7]. Therefore, these fields were designed to examine if changes in the reflex gain occur as part of the impedance controller which compensates for the environment. 2.3 Protocol All subjects practiced making movements in the NF on at least one day prior to the experiment. These training trials were used to accustom the subjects to the equipment and to the movement speed and accuracy requirements. Each force field was learned on a different day. Subjects first learned the DF, after which their endpoint stiffness was estimated. The order of training in the other two rotated DF force fields was randomized across subjects. For each force field, three procedures were conducted on the same day: learning, stiffness estimation, and after effects trials. During learning of the DF, subjects first completed 30 successful movements in the NF, after which the DF was activated, although no information was provided to the subjects about when activation of the DF would occur. Subjects then made movements in the DF until 100 successful trials were completed. For learning the rotated DF, they first performed 110 movements in the DF before making movements in the rotated DF until 100 successful trials were achieved. There was a short break before the stiffness estimation. Details of the method and analysis for the stiffness estimation can be found in [6,7,10]. Briefly, subjects first completed 20 successful trials in whichever force field was being tested. Following this, an additional 160 successful trials were completed in the force field of which 80 trials were randomly selected for stiffness estimation. During these trials, a 300ms displacement is applied near the midpoint of the trajectory. The displacement is divided up into three periods: 100ms ramp up, 100ms hold, and a 100ms ramp down period. The average force and displacement measured during the final 80 ms of the hold period were used to estimate the 2 by 2 endpoint stiffness matrix (
) by linear regression. Finally, after effects were recorded by randomly interspersing 20 null field trials with 80 force field trials. On these null field trials, subjects expect to be moving in the force field but no errors will be induced by the force fields. This allows for a clear picture of the feedforward control during the movements. On each experimental day, electromyographic activity of six arm muscles was measured. Surface EMG was recorded from two monoarticular shoulder muscles: pectoralis major and posterior deltoid; two biarticular muscles: biceps brachii and
Reflex Contributions to the Directional Tuning of Arm Stiffness 917 long head of the triceps; and two monoarticular elbow muscles: brachioradialis, and lateral head of the triceps. EMG was recorded using pairs of silver-silver chloride surface electrodes. The EMG signals were Butterworth band-pass filtered between 25 Hz and 1.0 kHz and sampled at 2.0 kHz. S1 S2 S3 S4 100 N/m A B C D -50
0 50 0 0.2 0.4
0.6 0.8
1 0 0.5 1 1.5
2 2.5
x 10 5 Orientation Shape Size
DF CW CCW DF CW CCW DF CW CCW
The stiffness ellipses for each of the four subjects participating in the study are shown for each of the three unstable force fields (DF: filled ellipse; CW DF: solid line; CCW DF: dotted line). The direction of the increased stiffness for each unstable force field is similar across all of the subjects. These directions are close to the directions of instability in the environment. B. The orientation of the endpoint stiffness ellipses. C. The shape of the endpoint stiffness ellipses. D. The size of the endpoint stiffness ellipses. Relative to the DF stiffness, the stiffness measured in the two rotated divergent force fields is larger and modified in orientation.
Subjects were able to adapt to all force fields and make smooth straight movements to the final target. While the onset of the force fields caused disturbed trajectories in early learning, subjects were able to reduce these errors gradually as learning progressed. After learning was completed, stiffness measurements were performed in each of the three unstable force fields. 3.1 Endpoint Stiffness The endpoint stiffness can be represented in terms of the eigenvalues and eigenvectors of the stiffness matrix using an ellipse. Singular value decomposition of the stiffness matrix K was used to find the eigenvalues [11]. Subjects modified the endpoint stiffness of their arms in different directions according to the different environments in which they were moving (Fig 2). Relative to the baseline stiffness in 918 G. Liaw et al. the 200 DF, the orientation of the endpoint stiffness in the two rotated DF fields had been changed. In each case the orientation occurred in the direction in which the environmental stiffness had been rotated. While the shape was fairly consistent across all fields, and more anisotropic than seen in null fields [7], the size of the stiffness was generally larger in the rotated DF fields. The endpoint stiffness can also be examined in terms of the four components of the stiffness matrix (Fig 3). For both the CW DF and the CCW DF, the K xx and K yy terms
were increased relative to the baseline DF level. However, the biggest difference in terms of the adaptation to the two force fields occurred in the K yx term. For the CW DF, this term was decreased relative to the baseline, whereas in the CCW DF this term had been increased. In contrast, the K xy term was maintained close to zero in both cases. Analysis of the effect of these terms was performed (Fig 3C). The change in the cross stiffness terms (K xy and K
yx ) was such that it compensates for the environmental instability with an oppositely directed force. 0 500 1000 0 500 1000 Kxx Kxy Kyx Kyy Kxx Kxy Kyx Kyy A B C force produced by Kxy force
produced by Kyx
perturbations in x & y axes CW DF CCW DF
CW DF CCW DF
Fig. 3. Changes in the stiffness relative to the DF stiffness. A,B. Mean endpoint stiffness across subjects. Each panel shows the mean values for one of the rotated DF fields relative to the stiffness in the baseline DF field (light grey bars). C. Effect of the change in cross terms of the endpoint stiffness matrix (K xy and K
yx ). For each of the rotated divergent force fields, the resulting change in the cross terms relative to the baseline DF stiffness was calculated. The force (solid line) resulting from a 1 cm displacement (dotted line) was calculated for each of the ΔK xy
yx terms. The K xy terms after adaptation produced only small changes in force. The ΔK yx produced opposite effects in each of the CW DF and CCW DF fields. These forces were directed to resist the oppositely directed forces from the divergent force field. 3.2 Electromyographic Activity The large difference between the two stiffness terms (K yx and K
xy ) (Fig 3) cannot normally be produced by pure co-contraction of opposing muscles [5, 7]. This difference, produced by modifying the antisymmetric part of the stiffness matrix, therefore suggests that subjects may have modulated part of their stiffness by
Reflex Contributions to the Directional Tuning of Arm Stiffness 919 Pectoralis Major +x +y -y -x -0.4
0 0.4
-0.4 0 0.4 -0.4 0 0.4 -0.4 0 0.4 -0.4 0 0.4 -0.4 0 0.4 -0.4 0 0.4 -0.4 0 0.4 DF CW CCW contrib. to stiffness [95-175 ms] long latency reflex
[60-125 ms] reflex response scaled for background activity
field. During the stiffness measurement the hand was moved by 8 mm in each of eight directions as illustrated by the arrow insert in the middle of the figure. The reflex response scaled by the background activity in unperturbed trials is shown for two intervals: a long latency reflex interval (60-125 ms) and the interval which could contribute to the stiffness measurement (95-175 ms). The response in each force field is plotted as separate bars: DF (left), CW DF (center) and CCW DF (right), with error bars denoting standard error of the mean. modifying the reflex gain. We investigated this by analyzing the reflex responses produced by the perturbations applied to measure stiffness. The integrated EMG was calculated during two periods after the perturbation: (60-125 ms) and (95-175 ms). The second interval was chosen as the period of muscle activation which would contribute to the stiffness estimate. This interval was determined by the fact that the stiffness was estimated between 120-200 ms after the onset of the perturbation and that the force produced by muscle activity (EMG) in the arm is time delayed by approximately 25 ms [12]. The integrated EMG was subtracted by the EMG normally present during this time in the movement (from unperturbed trials) and divided by the same number, producing a gain value which can be compared across the three fields.
920 G. Liaw et al. Posterior Deltoid +x +y -y -x 0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1 0 0.5 1 DF CW CCW contrib. to stiffness [95-175 ms] long latency reflex [60-125 ms] reflex response scaled for background activity
force field. During the stiffness measurement the hand was moved by 8 mm in each of eight directions as illustrated by the arrow insert in the middle of the figure. The reflex response scaled by the background activity in unperturbed trials is shown for two intervals: a long latency reflex interval (60-125 ms) and the interval which could contribute to the stiffness measurement (95-175 ms). The response in each force field is plotted as separate bars: DF (left), CW DF (center) and CCW DF (right), with error bars denoting standard error of the mean.
The reflex responses in the Pectoralis Major (Fig 4) demonstrate large differences in the long latency reflex responses depending on the force field to which the subject has adapted. For perturbations in the +x, +x –y, and –y directions, the reflex response has been inhibited for the CW DF field, and perhaps slightly enhanced in the CCW DF. For oppositely directed perturbations the opposite effect is seen. The CW DF shows an excitatory reflex response whereas the CCW DF is again inhibited. Similarly, the reflex responses in the Posterior Deltoid (Fig 5) show the same effect. The response of long latency reflex depends on the force field to which the subject has adapted. For perturbations in the +x, +x –y, and –y directions, the reflex response is excitatory for the CW DF field. For oppositely directed perturbations the
Reflex Contributions to the Directional Tuning of Arm Stiffness 921 opposite effect is seen. The CW DF shows an inhibited reflex response to the perturbations whereas the CCW DF is unaffected compared to the baseline field. Clearly changes in the reflex gain as part of the adaptation to the environment have occurred. Similar changes in the reflex responses were seen in all six arm muscles that were recorded. These reflex responses were modified appropriately to compensate for the destabilizing effects of the force fields.
Download 12.42 Mb. Do'stlaringiz bilan baham: |
ma'muriyatiga murojaat qiling