Eizo Miyashita

Optimal Feedback Control for Predicting Dynamic Stiffness During Arm Movement

Knowledge of the central nervous system (CNS) that constrains dimensions of freedom to control a redundant system of the body would provide inspiration for the robotic engineering. We estimated limb stiffness in Japanese monkeys (Macaca fuscata) during arm reaching movements using a robotic manipulandum and carried out numerical simulations. The estimated joint stiffness showed a diphasic pattern, and the end-point stiffness ellipses were modulated during the movement in a characteristic manner. The pattern of limb stiffness was reproduced by the numerical simulation using a musculoskeletal arm model and an approximate optimal feedback control (OFC). Although the arm model has a redundant system with multiple dimensions of freedom, the OFC minimized the redundancy by enhancing the task-relevant cost function. We suggest that the CNS may control the body according to a similar OFC law, and the knowledge might be useful for developing human-machine systems.

Eye movements following cortical stimulation in the ventral bank of the anterior ectosylvian sulcus of the cat

Eye movements were induced by stimulating the ventral bank of the anterior ectosylvian sulcus (AES) of chloralose-anesthetized cats. Intracortical microstimulation of this cortex evoked centering movements of both eyes. The latency of eye movement was 45 ms (range 40-60 ms) and the intensity of stimulation was 33 microA (range 20-45 microA). The evoked eye movements persisted after ablation of the classical frontal eye field of both sides. These results suggest that the ventral bank of the AES is involved in the control of eye movements.

Projections from eye movement-evoking cerebral cortices to the striatum and claustrum in the cat.

Projection fields in the striatum and claustrum from eye movement-evoking cerebral cortices (EMECs) in the cat were investigated using the WGA-HRP tracing method under electrophysiological guidance to place WGA-HRP in the appropriate sites. The EMECs in the frontal cortex, which are located in the medial wall under the cruciate sulcus (CRUo), the medial and lateral banks of the presylvian sulcus the the knee portion of the coronal sulcus (CORo), projected to the rostral part of the striatum. The EMEC in the temporal cortex, which is located in the ventral bank of the anterior ectosylvian sulcus, projected to the middle part of the striatum. Concerning the EMECs in the frontal cortex, the medially situated CRUo projected to the dorsomedial portion of the striatum, and the laterally situated CORo projected to the ventrolateral portion of the striatum. The dorsal half of the claustrum had reciprocal connections with all of the EMECs.

DEVISING A ROBOTIC ARM MANIPULANDUM FOR NORMAL AND ALTERED REACHING MOVEMENTS TO INVESTIGATE BRAIN MECHANISMS OF MOTOR CONTROL

The objective of this work was to develop a manipulandum system capable of measuring a monkeys arm movements and applying extrinsic forces to the hand. The manipulandum (RANARM) was designed as a five-bar linkage mechanism in order to realize sufficient rigidity and efficient force transmission. In addition, RANARMs control enables it to cancel its own dynamics by calculating the inverse dynamics. We evaluated the performance of RANARM from the mechanism, control, and application points of view. From the mechanism point of view, RANARM was able to provide a sufficiently large workspace for experiments, where it exhibited high manipulability and the capacity to apply force perturbations of at least 12 N under static conditions, and 5.3 N under dynamic conditions. From the control point of view, we estimated the uncompensated dynamics of RANARM to be negligibly small. Finally, we evaluated the performance of a monkey operating RANARM, and observed that the hand speed profiles followed a natural bell shape, while the maximal speeds were high (i.e., 15-cm reaching within 500 ms). In conclusion, RANARM allows a monkey to execute rapid movement and it is an efficient apparatus for investigating brain mechanisms for generating voluntary movements.

State Variables of the Arm May Be Encoded by Single Neuron Activity in the Monkey Motor Cortex

Revealing the type of information encoded by neurons activity in the motor cortex is essential not only for understanding the mechanism of motion control but also for developing a brain-machine interface. Thus far, the concept of preferred direction (PD) vector has dominated the discussion regarding how neural activity encodes information; however, a unified view of exactly what information is encoded has not yet been established. In this study, a model was constructed to describe temporal neuron activity by a dot product of the PD and the movement variables vector consisting of joint torque and angular velocity. The plausibility of this model was tested by comparing estimated neural activity with that recorded from the monkey motor cortex, and it was found that this model was able to explain the temporal pattern of neuron activity irrespective of its passive responsiveness. The mean determination coefficients of neurons that responded to proprioceptive stimuli and that responded to visual stimuli were relatively high values of 0.57 and 0.58, respectively. These results suggest that neurons in the monkey motor cortex encode state variables of the arm in a framework of modern control theory and that this information could be decoded for controlling a brain-machine interface.

A numerical simulation using optimal control can estimate stiffness profiles of a monkey arm during reaching movements

An understanding of how the brain constrains dimensions of freedom to control the body would be beneficial for the robotic engineering of a humanoid robot. We estimated joint stiffness in a female Japanese monkey (Macaca fuscata) during arm reaching movements and carried out a numerical simulation. The estimated stiffness was high at movement onset and movement end, and decreased at the mid-point of the movement. These characteristic patterns were reproduced by the numerical simulation using a 2-link 6-muscle arm model and an approximately optimal feedback control. Although the arm model was a redundant system with multiple dimensions of freedom, the optimal control was able to solve the redundancy problem by optimizing a task-relevant cost function. We suggest that the brain may control the body according to a similar optimal control law.

Suggestive evidence for a forward model of the arm in the monkey motor cortex

To test an idea that neurons in the motor cortex encodes a future state of the arm, we constructed a plausible model using arm state-related variables to explain neuronal activity, and applied a multiple linear regression analysis. We found that the model fit was fairly good with a mean determination coefficient of 0.57 for analyzed 231 neurons and that neuronal activity preceded the actual movement of the arm with 66 ms on average. Presuming that the brain follows optimal feed back control theory, these findings suggest that the motor cortex may contain a forward model of the arm.

Estimation of visual feedback contribution to limb stiffness in visuomotor control

The purpose of this work was to investigate contribution of a visual feedback system to limb stiffness. It is difficult to differentiate the visual component from others out of measured data obtained by applying a force perturbation, which is required to estimate stiffness,. In this study, we proposed an experimental procedure consisted of a pair of tasks to investigate the visual feedback component, and showed it as end-point stiffness ellipses at several timings of a movement. In addition, we carried out a numerical simulation of the movement with the perturbation in according with a framework of optimal feedback control model. As results, long axes of the stiffness ellipses of the visual component were modulated to the movement directions and the simulation showed that a positional feedback gain was exponentially increased toward a movement end. Consequently, the visual feedback system is supposed to regulate compliance of a movement direction.

Visual feedback effects as a coordination of joint stiffness in monkey's arm reaching

Hand or joint angle velocities correlated neuronal activity has been found in the monkey primary motor cortex during reaching arm movements. What variable does this activity represent in terms of control theory? Two types of control model regarding the reaching movement have been introduced in neuroscience; one is a control model that requires a desired trajectory (DTR) and the other is one that does not (DTnR). In DTR, the desired trajectory is input to both of parallelly arranged feedforward and feedback controllers. The feedforward controller produces a motor command through an inverse model and the feedback controller corrects errors by comparing the desired trajectory and an observed one. In DTnR, on the other hand, an initial state is sufficient to drive an optimal controller. The optimal controller produces a sequential motor command that is a multiplication of a state-feedback gain and an estimated state. A predicted state, which is produced by the motor command through a forward model, is used to produce the estimated state after adjustment by an observed state. Now, my question is that which of the variables the velocity-related neuronal activity represents: the desired trajectory, predicted state, estimated state, or motor command of a viscosity component. I propose two experiments. We may estimate the viscosity of an arm during the movement and evaluate a ratio of the viscosity component of the motor command to the total motor command. If the ratio is low enough, we may deny a possibility of the motor command component represented on the neuronal activity. To differentiate the rest of three variables, we may compare the neuronal activity with and without perturbation to an arm during the movement. The activity of the perturbed condition will be different from that of unperturbed condition if it represents the estimated state. We have to consider informational flow in addition to the experiments to differentiate the desired trajectory and predicted state.

Cocontraction of Pairs of Muscles around Joints May Improve an Accuracy of a Reaching Movement: a Numerical Simulation Study

We have pair muscle groups on a joint; agonist and antagonist muscles. Simultaneous activation of agonist and antagonist muscles around a joint, which is called cocontraction, is suggested to take a role of increasing the joint stiffness in order to decelerate hand speed and improve movement accuracy. However, it has not been clear how cocontraction and the joint stiffness are varied during movements. In this study, muscle activation and the joint stiffness in reaching movements were studied under several requirements of end‐point accuracy using a 2‐joint 6‐muscle model and an approximately optimal control. The time‐varying cocontraction and the joint stiffness were showed by the numerically simulation study. It indicated that the strength of cocontraction and the joint stiffness increased synchronously as the required accuracy level increased. We conclude that cocontraction may get the joint stiffness increased to achieve higher requirement of the movement accuracy.