JP5267413B2 - Operation assistance device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a movement assisting device capable of appropriately assisting the movement of a person. <P>SOLUTION: The movement assisting device includes an actuator for applying torque to the specified joint of a person, a storage unit, a sensor, and a controller for controlling the actuator. The sensor measures the movement pattern of a predetermined fixed part of the person. The storage unit stores a first movement pattern group describing each of the different movement patterns of the fixed part. Further, the storage unit stores target angle patterns for achieving each of the different second movement pattern group of the specified joint in association with each of the first movement pattern group. The controller selects the first movement pattern closest to the measured movement pattern from the first movement pattern group. The controller further controls the actuator so as to make the angle of the specified joint follow the target angle pattern associated with the selected first movement pattern. <P>COPYRIGHT: (C)2011,JPO&amp;INPIT

Description

  The present invention relates to a motion assisting device that assists human motion by applying torque to a joint.

  A walking assist device that assists walking motion by applying torque to the joints of the legs has been studied. For example, Patent Document 1 discloses a walking assistance device that assists walking of a person with one leg disabled. This walking assist device detects the motion of a healthy leg by a sensor, and applies torque to the joint of the non-free leg so that the motion of the non-free leg becomes the same as the motion of the normal leg.

JP 2006-314670 A

  When performing control to apply torque to a joint in order to assist human movement, some control target value is required for control of an actuator to apply torque. The technique of Patent Document 1 employs the detected motion pattern of a healthy leg as a control target value for an actuator that assists a non-free leg joint. Therefore, in the technique of Patent Document 1, if the motion of the healthy leg is not appropriate, the torque applied to the joint of the non-free leg becomes inappropriate for the walking motion. An appropriate control target value (target joint angle) for the joint to be controlled is necessary to appropriately assist not only the walking motion but also the human motion. Moreover, it is necessary to wear a brace on a healthy leg, which is a burden on the user. The present invention provides an operation assisting device capable of obtaining an appropriate control target value and appropriately assisting human movement.

  There are standardized movement patterns for the types of human movements. The walking motion is a motion that repeats a specific motion pattern, and is typical of stylized motion patterns. The operation of carrying the spoon into the mouth during meals and the repetitive work of factory workers are routine operations that repeat the same operation pattern. The inventor considered that a control target value (target angle pattern) for realizing an operation pattern can be prepared in advance if the operation is a fixed operation with a general pattern determined. However, for example, there may be a plurality of different motion patterns in detail even though the general motion patterns are the same, such as fast walking and slow walking. Therefore, the inventor prepares a plurality of target angle patterns for the joint to be controlled in advance, selects an appropriate one from a plurality of target angle pattern groups according to a human motion pattern, and adopts it as a control target value. I thought it would be good. In order to select an appropriate target angle pattern, a motion pattern of a predetermined part of a person is measured, and a target angle pattern group may be selected based on the measured motion pattern. The present invention was created based on such a concept. The “pattern” in this specification means a change with time of a certain parameter (such as a joint angle). That is, “pattern” has the same meaning as “time-dependent change”, “time-series data”, or “orbit”.

  The present invention can be embodied in an operation assisting device that assists human movement. The apparatus includes an actuator that applies torque to a specific joint of a person, a storage unit, a sensor, and a controller that controls the actuator. The “specific joint” means a joint to be controlled by the motion assisting device. The sensor measures an operation pattern of a predetermined part (predetermined part) of a person. For example, when the predetermined site is a joint, the sensor may be an angle sensor that detects the angle of the joint. The storage unit stores a first operation pattern group describing each of the different operation patterns of the predetermined parts. Further, the storage unit stores a target angle pattern for realizing each of the different second motion pattern groups of the specific joint in association with each of the first motion pattern groups. The controller selects the first operation pattern closest to the measured operation pattern from the first operation pattern group. The controller further controls the actuator so that the angle of the specific joint follows the target angle pattern associated with the selected first motion pattern.

  The motion assisting device selects the first motion pattern closest to the current actual motion pattern from the first motion pattern group stored in advance. “Closest” means that the pattern shapes are the most similar. A specific procedure for selection will be described later. Each first motion pattern is associated with a target angle pattern for realizing each of a plurality of second motion patterns including the movement of the specific joint. The motion assisting device identifies the target angle pattern associated with the selected first motion pattern and employs it as the control target pattern for the specific joint.

  An example of stored data will be described. Assume that the above-described motion assist device is applied to a walking assist device for a person with a knee joint of one leg (for example, the left leg). Assume that the hip joint of the left leg can be moved freely. In this case, for example, the hip joint of the left leg corresponds to the predetermined part, and the nonfree knee joint (failed knee joint) corresponds to the specific joint to be controlled. The first motion pattern corresponds to, for example, a motion pattern of the hip joint angle of the left leg. Each of the first operation pattern groups has a different walking speed and step length, for example. The second motion pattern corresponds to a motion pattern of a nonfree knee joint. That is, the target angle pattern of the failed knee joint is associated with each of the first motion patterns describing the temporal change in the hip joint angle of the left leg. For example, a target angle pattern of a failed knee joint for realizing an equivalent stride is associated with the movement of the hip joint angle of the left leg having a large stride (first motion pattern). Further, for example, a motion of the left leg hip joint angle (first motion pattern) having a high walking speed is associated with a target angle pattern of a failed knee joint for realizing an equivalent walking speed. This walking assist device selects the first motion pattern closest to the actual motion pattern of the left leg hip joint angle, and controls the failed knee joint so as to follow the target angle pattern associated with the first motion pattern. .

  In other words, the mechanism of the operation assisting device is as follows. The motion assisting device classifies whether the measured motion pattern of the predetermined part corresponds to a plurality of different second motion patterns prepared in advance. A target angle pattern of a joint to be controlled (specific joint) is associated with each classification. That is, the measured motion pattern of the predetermined part is classified, and the specific joint is controlled by the target angle pattern associated with the classification. With such a mechanism, the motion assisting device can cause the angle of the specific joint to follow the target angle pattern corresponding to the actual motion pattern of the predetermined part. Thus, the motion assisting device of the present invention can appropriately assist the specific joint.

  A description will be given of how to select the first operation pattern closest to the measured operation pattern from the first operation pattern group. Simply, it is conceivable to compare values at each time for the measured pattern and the first operation pattern. For example, it is conceivable to employ the root mean square of the difference in values for each time. However, since there are many points to be compared in the difference of the time values of the time series data, that is, the selection result may not be appropriate. It would be advantageous if the measured motion pattern and the characteristics of each first motion pattern could be represented with a smaller number of parameters. This is because it is only necessary to compare the difference in the parameter values representing the characteristics of the measured motion pattern and the first motion pattern. By narrowing down the number of parameters to be compared, the closest first operation pattern can be accurately selected. One of the techniques disclosed in this specification uses a Fourier transform to narrow down parameters when comparing operation patterns.

  The time series data can be reduced in parameters representing the characteristics of the time series data by performing Fourier transform to obtain a Fourier coefficient set. This is because the low-frequency Fourier coefficient expresses the characteristics better than the high-frequency Fourier coefficient, and the time-series data characteristics can be expressed with the low-frequency Fourier coefficients ignoring the high-frequency Fourier coefficients. is there.

  The inventor applied this knowledge to the selection of the first operation pattern. Specifically, the motion assisting device preferably has the following configuration. The storage unit stores a first Fourier coefficient set obtained by Fourier transforming time-series data of the motion pattern of the predetermined part indicated by each first motion pattern. Then, the controller calculates a second Fourier coefficient set obtained by Fourier transforming the time series data of the measured motion pattern. The controller compares the second Fourier coefficient set with each first Fourier coefficient set, and selects the first motion pattern corresponding to the first Fourier coefficient set having the smallest difference from the second Fourier coefficient set. For example, the feature of an operation pattern composed of 1000 points of time-series data can be represented by, for example, 10 low-frequency Fourier coefficients obtained by Fourier transform. In this case, the number of parameters representing features can be reduced from 1000 to 10. For example, in this case, the controller may select the first motion pattern in which the difference between the ten Fourier coefficients of the measured motion pattern and the ten Fourier coefficients representing each first motion pattern is the smallest. More preferably, the controller may select the first operation pattern corresponding to the first Fourier coefficient set having the smallest mean square error with the second Fourier coefficient set. By adopting the Fourier coefficient, the first operation pattern close to the measured operation pattern can be selected at high speed and accurately.

  ADVANTAGE OF THE INVENTION According to this invention, the operation assistance apparatus which can assist a person's operation | movement appropriately is implement | achieved.

The schematic diagram of a movement assistance apparatus is shown. It is a figure which shows the example of a 1st operation | movement pattern. It is a figure which shows the example of a 1st Fourier coefficient set. It is a figure which shows the example of a target angle pattern. It is a flowchart figure of a control process. It is a flowchart figure of an update process. It is a block diagram of the controller of the operation assistance apparatus of 2nd Example. A schematic diagram of a user's leg wearing a leg brace is shown. It is a graph of the result of the simulation in 2nd Example (1). It is a graph of the result of the simulation in 2nd Example (2).

  Some of the technical features of the operation assistance device of the embodiment will be listed.

  (First feature) When the minimum difference between the second Fourier coefficient set and each first Fourier coefficient set is larger than a predetermined threshold value, the controller changes the command to the actuator and continues the control. For example, only feedback control that emphasizes that a person does not fall is executed. The advantages of such a configuration are as follows. The measured motion pattern is not necessarily close to any of the first motion pattern groups prepared in advance. The case where the minimum value of the difference between the second Fourier coefficient set and each first Fourier coefficient set is larger than the predetermined threshold corresponds to the case where the first operation pattern does not correspond to any of the first operation patterns prepared. .

  (Second feature) The controller stores the measured motion pattern in the storage unit in association with the selected first motion pattern. Further, the controller updates the first motion pattern by the weighted average of the newly stored measured motion pattern and the past measured motion pattern group stored in association with the first motion pattern. This process corresponds to updating the first motion pattern using the “measured motion pattern” classified into the first motion pattern. Incorporating such processing, so-called “learning” in the technical field of control is performed, and a first operation pattern adapted to each user is formed.

  (Third feature) The weighted average weight is preferably a forgetting factor that decreases as the elapsed time from the stored time becomes longer. By incorporating such processing, the degree of contribution of old data (measurement operation pattern stored in the past) to the formation of the first operation pattern can be reduced.

  (Fourth feature) When the number of past measured motion patterns associated with the first motion pattern is larger than a predetermined value, the controller stores higher-order Fourier coefficients than before.

  (Fifth feature) When the speed of the predetermined part detected by the sensor is larger than the threshold, the sampling time when the sensor measures the motion pattern of the predetermined part is shortened. The higher the speed of the predetermined part, the higher the frequency is included in the operation pattern. Therefore, by shortening the sampling time, it is possible to capture a high frequency component included in the operation pattern of the predetermined part.

  A motion assisting device 10 according to a first embodiment will be described with reference to the drawings. The motion assisting device 10 of the present embodiment is a device that assists the user's walking motion by applying an appropriate torque to the user's left knee joint. In this embodiment, it is assumed that the user cannot freely move the left knee joint. However, it is assumed that the left hip joint (joint that swings the thigh about the pitch axis) can be freely moved. FIG. 1 shows a schematic diagram of the motion assisting device 10 worn by the user. 1A shows a front view, and FIG. 1B shows a side view.

  The motion assisting device 10 includes a leg brace 12L that is worn along the left leg of the user.

  The left leg orthosis 12L will be described in detail. The leg orthosis 12L is attached to the outside of the left leg from the user's thigh to the lower leg. The leg orthosis 12L has a multi-link mechanism having an upper link 14L, a lower link 16L, and a sole link 18L. The upper end of the upper link 14L is connected to the support link 30 via the waist joint 20aL. The lower link 16L is connected to the upper link 14L by a knee joint 20bL located outside the knee. The sole link 18L is swingably coupled to the lower link 16L by an ankle joint 20cL located outside the user's ankle. The upper link 14L is fixed to the user's thigh with a belt. The lower link 16L is fixed to the user's lower leg with a belt. The sole link 18L is fixed to the user's sole with a belt. The belt for fixing the sole link 18L is not shown. The support link 30 is fixed to the trunk (waist) of the user.

  When the user wears the leg brace 12L, the waist joint 20aL, the knee joint 20bL, and the ankle joint 20cL are positioned coaxially with the pitch axis of the user's hip joint, the pitch axis of the knee joint, and the pitch axis of the ankle joint, respectively. . That is, the leg orthosis 12L can swing according to the movement of the user's left leg. An encoder 21 for detecting the angle between the links is attached to each joint. In the following, the angle between the links is referred to as the joint angle. The encoder group 21 attached to each joint is generically referred to as a leg sensor 21.

  A motor 32 (actuator) is attached to the knee joint 20bL. The motor 32 is located outside the user's knee joint. The motor 32 can rotate the lower link 16L with respect to the upper link 14L. That is, the motor 32 can apply torque to the user's left knee joint. Although not shown, the motor 32 includes a clutch. When the clutch is released, the torque transmission path from the motor 32 to the left knee joint is interrupted. When the clutch is released, the user can freely swing the left knee joint.

  Reference numeral 40 in the figure indicates a controller that controls the motor 32. The leg sensor 21 included in the leg brace 12L corresponds to a sensor that measures a leg motion pattern.

  The controller 40 has a built-in storage unit (not shown) and stores various data. The stored data will be described. The storage unit stores an operation pattern (time series data) of the left hip joint angle during walking. Hereinafter, the motion pattern of the left hip joint angle is referred to as a first motion pattern. The storage unit stores a plurality of different first operation patterns. The storage unit stores a first Fourier coefficient set obtained by Fourier transforming time series data of each operation pattern. The first Fourier coefficient set is composed of 10 Fourier coefficients in order from the lowest frequency. Further, the storage device stores a plurality of different left knee joint target joint angle patterns (target angle patterns). Each target joint angle pattern is associated with each first motion pattern.

  The data stored in the storage unit will be described in detail with reference to FIGS. FIG. 2 is a schematic graph showing some first operation patterns. The vertical axis is the left hip joint angle, and the horizontal axis is time. FIG. 2 shows three first operation patterns RPT1, RPT2, and RPT3. Although there are more than three stored first operation patterns, only three first operation patterns are illustrated in FIG. 2 for convenience. The symbol TS in FIG. 2 indicates one walking cycle. That is, the first motion pattern group is a set of different motion patterns in one walking cycle. For example, the first operation pattern RPT2 corresponds to an operation pattern having a larger step than the first operation pattern RPT1. The first operation pattern RPT3 corresponds to an operation pattern having a faster walking speed than the first operation pattern RPT1.

  FIG. 3 shows a schematic graph of the first Fourier coefficient set corresponding to each of the first operation patterns. The horizontal axis indicates the frequency, and the vertical axis indicates the amplitude of each frequency component. The first Fourier coefficient set FQ1 is data obtained by Fourier transforming the data of the first operation pattern RPT1. The first Fourier coefficient set FQ2 is data obtained by Fourier transforming the data of the first operation pattern RPT2. The first Fourier coefficient set FQ3 is data obtained by Fourier transforming the data of the first operation pattern RPT3. Each first Fourier coefficient set is composed of 10 Fourier coefficients. For convenience, FIG. 3 shows only 4 coefficients from the low frequency side. For example, the coefficient w11 represents the amplitude of the primary frequency component of the first operation pattern RPT1. Similarly, the coefficients w12, w13, and w14 represent the amplitudes of the second, third, and fourth frequency components of the first operation pattern RPT1, respectively.

  FIG. 4 is a graph schematically showing a target angle pattern of the left knee joint angle used when the motor 32 is controlled. The horizontal axis represents time, and the vertical axis represents the target angle of the left knee joint. Each target angle pattern is associated with each of the first operation patterns described above. FIG. 4 shows three target angle patterns TPT1, TPT2, and TPT3. Although there are more than three target angle patterns stored, FIG. 4 illustrates only three target angle patterns for convenience.

  A symbol TS in FIG. 4 indicates one walking cycle. Each of the target angle patterns corresponds to a control target angle pattern of the left knee joint angle for realizing different motion patterns of one walking cycle of the left leg. The motion pattern of one walking cycle of the left leg corresponds to the second motion pattern.

  The three target angle patterns TPT1, TPT2, and TPT3 are associated with the first operation patterns RPT1, RPT2, and RPT3, respectively. It should be noted that the graphs of FIGS. 2 and 4 are schematic representations of changes in joint angles over time, and do not specifically show changes in actual joint angles over time. .

  The control process executed by the controller 40 will be outlined. The controller 40 measures the motion pattern of the left hip joint angle using the leg sensor 21. Next, the controller 40 selects an operation pattern closest to the measured operation pattern from the stored first operation pattern group. The controller 40 controls the motor 32 so that the angle of the user's left knee joint follows the target angle pattern associated with the selected first motion pattern. In this embodiment, the left hip joint of a person corresponds to a “predetermined part”, and the left knee joint to be controlled corresponds to a “specific joint”.

  With reference to FIG. 5, the control process of the motor 32 which the controller 40 performs is demonstrated concretely. FIG. 5 is a flowchart of the control process. The controller 40 measures the left hip joint angle by the leg sensor 21 (S2). When the measurement data for a certain time is accumulated, that is, when the motion pattern of the left hip joint angle for a certain time is measured, the controller 40 performs a fast Fourier transform on the measured motion pattern to obtain a Fourier coefficient set (S4). The fast Fourier transform algorithm is well known and will not be described. The acquired Fourier coefficient set is referred to as a second Fourier coefficient set.

  Next, the controller 40 calculates the root mean square error of the difference between the acquired second Fourier coefficient set and each of the stored first Fourier coefficient set groups (S6). Specifically, the square of the difference between the n-th frequency component of the second Fourier coefficient set and the second frequency component of the first Fourier coefficient set is added by 10 frequency components and divided by 10. By the process in step S6, the same number of root mean square errors as the number of stored first operation patterns is obtained.

  Next, the controller 40 selects the first operation pattern with the smallest mean square error (S8). By the processing in step S8, the first motion pattern corresponding to the first Fourier coefficient set closest to the second Fourier coefficient set of the measured motion pattern is selected. Next, the controller 40 compares the minimum value of the mean square error with a predetermined threshold value (S10). When the minimum value of the mean square error is larger than a predetermined threshold (S10: NO), the controller 40 changes the command generation rule for the motor and continues the control (S16). As the rule after the change, for example, feedback control that emphasizes that a person does not fall down is adopted. As described above, when the minimum value of the mean square error is larger than a predetermined threshold value, it means that the measured motion pattern does not correspond to any of the stored first motion pattern groups. To do. This motion assisting device 10 is a feedback control that places importance on the fact that the control rule up to that point does not fall down if the measured motion pattern does not correspond to any of the stored first motion pattern groups. Switch to.

  When the minimum value of the mean square error is smaller than a predetermined threshold value (S10: YES), the controller 40 specifies a target angle pattern corresponding to the selected first motion pattern (S12). Next, the controller 40 controls the motor 32 so that the angle of the left knee joint follows the specified target angle pattern (S14). The motor 32 applies torque to the left knee joint to assist the user's walking motion.

  The advantages of the control process described above will be described. The measured motion pattern is slightly different for each walking cycle. However, the variation of the second Fourier coefficient set obtained by Fourier transforming the measured motion pattern data is smaller than the slight variation of the motion pattern for each walking cycle. This is because the second Fourier coefficient set mainly represents the global characteristics of the operation pattern. Therefore, the control process described above can suppress the influence of slight fluctuations in the motion pattern for each walking cycle, and determine which of the first motion pattern groups in which the measured motion pattern is stored is closest. Therefore, the above-described control process can assist the walking motion with a torque suitable for the user's motion pattern.

  Next, processing executed by the controller 40 in parallel with the control processing will be described. In parallel with the motor control, the controller 40 executes a process for updating the first operation pattern stored in the measured operation pattern. FIG. 6 shows a flowchart of the update process. The controller 40 stores the motion pattern (time series data of the left hip joint angle) measured in step S2 described above in association with the selected first motion pattern (S20). By the process of step S20, a plurality of past measurement operation patterns are associated with the specific first operation pattern. The controller 40 calculates a weighted average of past measurement operation patterns associated with the specific first operation pattern (S22). The controller 40 updates the first operation pattern with the calculated weighted average (S24). The formula for calculating the weighted average is shown in the following (Equation 1).

  In (Equation 1), Pθ (t) represents the knee joint angle of the first motion pattern. t is a variable indicating the timing in the time span of one walking cycle. Siθ (t) represents the knee joint angle of the i-th past measurement operation pattern. n represents the number of stored measurement operation patterns. As the index i is smaller, Siθ (t) indicates older measurement data. The coefficient ai indicates a weighting coefficient for calculating the weighted average, and is given by the following (Equation 2).

  (Equation 2) indicates that the older the data, the smaller the influence that contributes to the weighted average. That is, the coefficient ai functions as a forgetting factor in the technical field of time series analysis.

  The update process of FIG. 6 has the following meaning. Each of the stored first operation patterns is represented by a weighted average of past measurement operation patterns classified into the first operation pattern. Each time an operation pattern is measured, the first operation pattern is updated by incorporating the feature. The degree of contribution to the updated first operation pattern becomes smaller as the measurement operation pattern is older. That is, the update process in FIG. 6 means that the motion assisting device 10 updates (learns) the stored first motion pattern in accordance with the latest pattern of the user's walking motion. By this update process, the motion assisting device 10 can assist the walking motion by applying an appropriate torque to the broken left knee joint while learning the user's walking motion.

  Points to be noted in the processing of the controller 40 will be described. The controller 40 calculates the rotational angular velocity of the left hip joint during the process of S2. The rotational angular velocity is obtained as a time difference of the time series data of the left hip joint angle. The controller 40 shortens the sampling time for measuring the left hip joint when the maximum value of the rotational angular velocity is larger than a prescribed threshold value. This process can be applied to a case where a user's operation pattern has a large time change, and a higher frequency component can be acquired.

  Further, when the number of past measurement operation patterns stored in the respective first operation patterns is larger than the predetermined number, the controller 40 stores higher-order Fourier coefficients than before. If the number of past measurement operation patterns is large, there is a high possibility that the operation will be repeated again. By storing higher order Fourier counts for such motion patterns, the motion patterns can be accurately reproduced.

  The points to be noted of the motion assisting device 10 will be described. In the movement assisting device 10, the left hip joint of a person whose movement pattern is measured by a sensor corresponds to the predetermined part. The part to be measured is not limited to the angle of the left hip joint. For example, the motion assisting device 10 may measure the trajectory of the right leg foot. In this case, the first motion pattern corresponds to data describing the trajectory (ie motion pattern) of the right leg foot.

  A second embodiment will be described. The second embodiment is a motion assisting device that measures the motion pattern of the user's left leg and applies torque to the left knee joint based on the measured motion pattern. The equipment configuration of this motion assisting device is the same as the leg brace 12L of the motion assisting device 10 of the first embodiment shown in FIG. Accordingly, FIG. 1 is also referred to in the description of the motion assisting device of the second embodiment. However, the leg brace of the second embodiment is referred to as a leg brace 107. The motion assisting device of the second embodiment includes a controller 100 and leg braces 107. The configuration of the leg brace 107 is the same as that of the leg brace 12L in FIG. However, the leg orthosis 107 also includes a motor in the ankle joint 20cL. Detailed description of the structure of the leg brace 107 is omitted.

  FIG. 7 shows a block diagram of the controller 100 of the motion assist device of the second embodiment. The device indicated by each symbol in FIG. 7 will be described. Reference numeral 101 denotes a Fourier coefficient memory, reference numeral 102 denotes a Fourier coefficient calculator, reference numeral 103 denotes a motion intention estimator, reference numeral 104 denotes a target motion generator, and reference numeral 105 denotes each joint command generator. 106 indicates a controller, 107 indicates a leg brace, 109 indicates a state detector, and 110 indicates a sample time adjuster. Reference numeral 108 denotes a user of the motion assist device.

  The motion trapping device according to the present embodiment is intended for a person who cannot move freely from the knee of one leg. The other leg is healthy. The leg on the side that cannot be freely moved is hereinafter referred to as a non-healthy leg.

  The Fourier coefficient storage unit 101 stores Fourier coefficient calculation values, which are Fourier coefficients of past state detection value adjustment values, and outputs them as Fourier coefficient storage values as necessary. The Fourier coefficient calculation value to be stored is input from the Fourier coefficient calculator 102. A state detection value adjustment value is input to the Fourier coefficient calculator 102. The Fourier coefficient calculator 102 calculates a Fourier coefficient based on the input value and outputs it as the Fourier coefficient calculation value described above. The motion intention estimator 103 receives the Fourier coefficient stored value and the Fourier coefficient calculation value. The motion intention estimator 103 estimates and outputs the motion intention of the user 108 based on a plurality of values closest to the Fourier coefficient calculation value among the Fourier coefficient stored values.

  The target motion generator 104 receives the motion intention as an input. The target motion generator 104 generates and outputs a target motion that is a desired motion of the leg brace 107 worn by the user 108 based on the input signal. Each joint command generator 105 is supplied with the above-described target motion, the above-described state detection value, and the above-described Fourier coefficient storage value as inputs, and each joint command generator 105 is a desired angle of each joint of the leg orthosis 107 based on the input signal. Generate and output a joint command. The controller 106 is supplied with the above-described joint commands and the above-described state detection values as inputs, and drives the leg brace 107 with a torque command such that the above-described state detection values follow the above-described joint commands.

  The leg brace 107 assists the operation of the user 108 to wear, and each joint angle is detected by the state detector 109 and output as the above-described state detection value. The sample time adjuster 110 adjusts the sample time of the aforementioned state detection value and outputs it as the aforementioned state detection value adjustment value.

  FIG. 8 is a schematic diagram of a user's leg wearing the leg brace 107. Reference numerals in FIG. 8 will be described. Reference numeral 201 indicates a trunk, reference numeral 202 indicates a thigh, reference numeral 203 indicates a knee, reference numeral 204 indicates a shin, reference numeral 205 indicates an ankle, reference numeral 206 indicates a foot, reference numeral 207 indicates a healthy leg. Reference numeral 208 denotes the ground. In addition, θh represents the unhealthy leg thigh angle, θk represents the unhealthy leg knee angle, θa represents the unhealthy leg ankle angle, θi represents the healthy leg thigh angle, and lg represents the unhealthy leg. The thigh length is indicated, and ln indicates the non-healthy leg shin length.

  In the present embodiment, an example of a leg brace that assists a user who cannot move freely below the knee of one leg will be described. In FIG. 8, a leg brace 107 is attached to each part of the non-healthy leg thigh 202, the non-healthy leg knee 203, the non-healthy leg shin 204, the non-healthy leg ankle 205, and the non-healthy leg foot 206 shown in bold lines Suppose that the non-healthy leg knee 203 and the non-healthy leg ankle 205 are assisted by using an actuator. Further, it is assumed that the user 108 can move the non-healthy leg thigh angle θh and the healthy leg 207 with his / her own intention. At the moment shown in the figure, the unhealthy leg is a standing leg, supports the trunk 201 of the user 108, and walks on the ground 208. A case where the healthy leg 207 is not provided with a detector will be described.

  Hereinafter, the details of the mechanism by which the controller 100 of the present embodiment drives the leg brace 107 will be described using the simulation results of FIGS. 9 and 10.

  The numerical values used in this simulation are as follows. m = 45 [kg], mk = 10 [kg], ma = 5 [kg], lg = 0.45 [m], ln = 0.45 [m], lf = 0.2 [m], g = 9.8 [m / s ^ 2], T = 10 × 10 ^ −3 [s], Jk = (2 * m + mk + ma) * lg ^ 2, Ja = (2 * m + 2 * mk + ma) * ln ^ 2, Kpk = 10 [s ^ -1], Kvk = Kpk · (2π) [s ^ -1], Kvjk = Kvk · Jk [N · m · s / rad], Kpa = 10 [s ^ -1], Kva = Kpa · (2π) [s ^ −1], Kvja = Kva · Ja [N · m · s / rad], vh = −v / (lg + ln) [rad / s], θkmin = −175 · π / 180 [ rad], θkmax = −0.5 · π / 180 [rad], θhmin = −60 · π / 180 [rad], θhmax = 90 · π / 180 [rad], θimin = θhmin rad], θimax = θhmax [rad], ymin = 0.8 · (lg + ln) [m], ymax = 0.9 · (lg + ln) [m], N0 = 10, λ = 0.01.

  Where m is the trunk mass, mk is the knee mass, ma is the ankle mass, lg is the thigh length, ln is the shin length, lf is the foot length, g is the gravitational acceleration, T is the control period, and Kpk is the knee angle. Proportional control gain, Kvk is normalized knee angular velocity proportional control gain, Kvjk is knee angular velocity proportional control gain, Kpa is ankle angle proportional control gain, Kva is normalized ankle angular velocity proportional control gain, Kvja is ankle angular velocity proportional control gain, and vh is Thigh angular velocity command, v is walking speed, θkmin is knee angle lower limit value, θkmmax is knee angle upper limit value, θhmin is thigh angle lower limit value, θhmax is thigh angle upper limit value, θimin is swing leg lower limit value Ymin represents the trunk vertical position command lower limit value, ymax represents the trunk vertical position command upper limit value, N0 represents the fundamental order of Fourier transform, and λ represents the forgetting factor. Here, the basic order N0 and the forgetting factor λ are parameters used in the Fourier coefficient storage unit 101, and details thereof will be described later.

  In this simulation, first, biped walking using only feedback control is performed with walking speeds v = 0.5 [m / s], v = 1.0 [m / s], and v = 1.5 [m / s]. Do this for each, and create motion intentions that are different walking patterns. Here, the ground is horizontal, both legs have the same mechanical constant, and each joint angle and ankle angle of the free leg are determined based on actual human walking. In the feedback control described above, the knee and ankle were set to position proportional speed proportional control with the above control parameters set, the trunk vertical position was between the upper limit value and the lower limit value, and walking was performed at a desired walking speed. Here, the ankle angle was a function of the thigh angle, and the relational expression was obtained based on walking actual measurement data. Data from Dr. Ben Stansfield, University of Strathclyde was used as walking measurement data. However, when the user 108 actually uses the leg orthosis 107, the normal leg and the non-healthy leg thigh are operated by the user 108 by his / her own intention, and the non-healthy leg knee angle and the non-healthy leg ankle angle only. Are driven by the above feedback control.

  FIG. 9 is an example of the operation intention showing the second embodiment obtained by the simulation. FIGS. 9A, 9B, and 9C show the cases of walking speeds of 0.5 [m / s], 1.0 [m / s], and 1.5 [m / s] (hereinafter “No. 1 operation intention ”,“ second operation intention ”, and“ third operation intention ”). In the figure, the alternate long and short dash line is the thigh angle, the broken line is the knee angle, the solid line is the ankle angle, t1 is the time when the thigh speed is positive and the thigh angle is 0 [rad], and t2 is the thigh The time when the thigh angle is maximum, t3 is the time when the thigh speed is negative and the thigh angle is 0 [rad], and t4 is the time when the thigh angle is minimum. In FIG. 9A, t2 = t1 + 1.2 [s], t3 = t2 + 2.0 [s], t4 = t3 + 1.5 [s], and t5 = t4 + 0.9 [s]. 9B, t2 = t1 + 0.8 [s], t3 = t2 + 1.0 [s], t4 = t3 + 0.7 [s], t5 = t4 + 0.4 [s], and in FIG. t2 = t1 + 0.6 [s], t3 = t2 + 0.5 [s], t4 = t3 + 0.4 [s], and t5 = t4 + 0.3 [s].

  The Fourier coefficient calculator 102 in FIG. 7 calculates Fourier coefficients in the sections [t1, t2], [t2, t3], [t3, t4], [t4, t5] of the thigh angle θh in FIG. ).

  However, the thigh angle θh is adjusted by the sample time adjuster 110 so that it becomes Mk point data at equal intervals in each of the above sections, and its β-th value is θhβ. Further, fhα is a Fourier coefficient calculation value calculated by a discrete Fourier transform of θhβ. The Fourier coefficient calculator 102 similarly calculates Fourier coefficients for the knee angle θk and the ankle angle θa, the thigh angle Fourier coefficient fhα, the knee angle Fourier coefficient calculated value fkα, and the ankle angle Fourier coefficient calculated value faα Fourier-transformed. It outputs to the Fourier coefficient memory | storage device 101 and the motion intention estimation device 103 as a coefficient calculation value.

  The Fourier coefficient storage unit 101 stores values up to the Nk order among the Fourier coefficient calculation values fhα. However, the order of the Fourier coefficient to be stored is determined based on (Equation 4).

  Here, Nk is the order of the Fourier coefficient of the waveform in the kth section to be stored, N0 is the basic order, λ is the forgetting coefficient, n is the current section, and r is the number of times the same operation is repeated. The basic order N0 is set according to the memory capacity mounted on the leg brace 107. When the forgetting factor λ is increased, the Fourier coefficient stored value based on the past walking experience is quickly erased (forgotten) from the Fourier coefficient memory 101. Further, as the same walking pattern is repeated, the aforementioned Fourier coefficient is stored up to a higher order, and as the walking pattern performed at the present time, the aforementioned Fourier coefficient is stored up to a higher order. By setting the order of Fourier coefficients to be stored as in (Equation 4), only the Fourier coefficients of the walking patterns that are more likely to be used are preferentially stored, so the memory usage can be saved and the cost can be reduced. The leg brace 107 can be configured, and the motion intention estimation can be performed in a short time.

  In the sample time adjuster 110, based on the thigh speed dθh / dt at the point where the negative sign of the thigh angle θh changes and the thigh position θh when the thigh speed dθh / dt changes, The walking speed is estimated, and the fundamental frequency of the Fourier transform is set higher in proportion to the walking speed. That is, as shown in (Equation 5), the number of sample points Mk in each section is set large.

  However, l is a motion intention number (for example, l = 2 indicates a second motion intention), v (lk) is the walking speed in the walk of the first lk motion intention, c is a parameter for determining the number of sample points Mk, round (・) Is a function (round function) that rounds off the decimal point, dθh + / dt is the thigh speed when the thigh position θh cuts 0 in the positive direction, and dθh + (l) / dt is large in the intention of the l-th motion. The thigh speed when the thigh position θh cuts 0 in the positive direction, dθhmax / dt is the thigh speed when the thigh position θh takes the maximum value, and dθhmax (l) / dt is the l-th operation. The maximum value of the thigh position θh in the intention, dθh− / dt is the thigh speed when the thigh position θh cuts 0 in the negative direction, and dθh− (l) / dt is large in the l-th motion intention. The thigh when the thigh position θh cuts 0 in the negative direction Time, dθhmin / dt is thigh speed when the thigh position θh takes a minimum value, dθhmin (l) / dt is the minimum value of the thigh position θh in the l operation intended.

  The Fourier coefficient storage unit 101 stores the Fourier coefficient calculation values described above as Fourier coefficient storage values up to the Nk order shown in (Equation 4). Hereinafter, the thigh angle Fourier coefficient stored value of the non-healthy leg stored in the Fourier coefficient storage unit 101 from the past motion, and the non-healthy leg size obtained in the same manner as described above while the user 108 is walking. A mechanism for estimating the motion intention of the user 108 based on the thigh angle Fourier coefficient calculation value fhα and controlling the unhealthy leg knee angle and the unhealthy leg ankle angle will be described. Here, the movement of the healthy leg is the one in which the three walking patterns in FIG. 9 continue in a random order, the walking pattern of the healthy leg is in the same order as the non-healthy leg thigh angle, and the time is It is assumed that each walking pattern is half a cycle earlier than the time of the healthy thigh angle.

  FIG. 10 shows simulation results of motion intention estimation according to the second embodiment. In the figure, a one-dot chain line indicates a non-healthy thigh angle, a broken line indicates a non-healthy knee angle, and a solid line indicates a non-healthy leg ankle angle. Also, t1 = 0 [s], t2 = 1.2 [s], t3 = 3.2 [s], t4 = 3.9 [s], t5 = 4.4 [s], t6 = 5.1 [s], t7 = 6.0 [s], t8 = 6.4 [s], t9 = 6.8 [s], t10 = 7.5 [s], t11 = 8.5 [s], t12 = 9.1 [s], t13 = 9.7 [s], and t14 = 10.7 [s]. Hereinafter, each part operation | movement of FIG. 7 is demonstrated for every area.

  It is assumed that the walking pattern shown in FIG. 10 has already been experienced once at time t1, and the Fourier coefficient storage unit 101 stores a Fourier coefficient storage value based on the walking experience.

  In section [t1, t2], there is no information on past non-healthy leg thigh angles, so the current motion intention estimation cannot be performed, and the non-healthy knee angle and the non-healthy ankle angle are implemented by the feedback control. ing. The Fourier coefficient calculator 102 calculates a Fourier coefficient calculation value of the non-healthy thigh angle and stores it in the Fourier coefficient storage 101. The Fourier coefficient storage unit 101 outputs the Fourier coefficients of the three walking patterns shown in FIG. 9 stored from past walking experiences as Fourier coefficient storage values. The motion intention estimator 103 compares the Fourier coefficient calculated value and the Fourier coefficient stored value by (Equation 6), and determines that the Fourier coefficient calculated value is closest to the Fourier coefficient stored value of the first walking pattern. . As a result, it is determined that the user 108 intends to walk at a walking speed of 0.5 [m / s] shown in FIG. 9A (hereinafter referred to as “first motion intention”), and this is indicated. Output as motion intention.

  However, l is a motion intention number (for example, l = 2 indicates a second motion intention), and δ (l) is a current Fourier coefficient stored value fα and a Fourier coefficient stored value fα (l) of the lth motion intention. The Fourier coefficient stored value error sum of squares lk, which is the sum of the squares of the differences, lk is an operation intention number l that minimizes the Fourier coefficient stored value error sum of squares δ (l). Here, since the Fourier coefficient storage value fα and the Fourier coefficient storage value fα (l) intended for the l-th operation are generally different orders (Nk ≠ Nk (l)), the Fourier coefficient storage value in (Expression 6). The error sum of squares δ (l) is the sum up to the smaller order.

  The target motion generator 104 inputs the motion intention, and “moves horizontally at about 0.5 [m / s]” common to the first motion intention and “the vertical position of the trunk 201 is 0”. .. Keeping 75 [m] or more "is output.

  Each joint command generator 105 is supplied with the target motion, the state detection value, and the Fourier coefficient stored value as input values, and the horizontal movement speed and the vertical position calculated based on the state detection value are the upper target motion. Is within the set error range, the inverse Fourier transform is applied to the Fourier coefficient stored value of the first walking pattern, and the non-healthy leg knee angle command and non-healthy calculated by (Equation 8) An ankle angle command is output as each joint command, and when the horizontal movement speed calculated based on the state detection value and the vertical position are not within the set error range from the target motion, the target motion is satisfied. A non-healthy knee angle command and a non-healthy leg ankle angle command are output as the joint commands. This simulation is the former case.

  However, θkk is a non-healthy leg knee angle command when the horizontal movement speed and the vertical position calculated based on the state detection value are within the set error range from the target motion, and fkα is stored in the Fourier coefficient storage unit 101. The knee angle Fourier coefficient of the k section to be performed, θak represents a non-healthy leg ankle angle command when the horizontal movement speed and the vertical position calculated based on the state detection value are within the set error range from the target motion. . Further, faα represents the ankle angle Fourier coefficient in the k section stored in the Fourier coefficient storage unit 101, and Mk represents the number of sample points in the k section.

  In the section [t2, t3], the controller 106 receives the joint commands and the state detection values as inputs. When the horizontal movement speed calculated based on the state detection value and the vertical position are within a set error range from the target motion, the controller 106 causes the state detection value to follow each joint command. A torque command that is a feedforward control command is output. The controller 106 is a feedback control command that satisfies the target operation when the horizontal movement speed and the vertical position calculated based on the state detection value are not within the set error range from the target operation. Outputs a torque command.

  Actuators mounted on the non-healthy leg knee 203 and the non-healthy leg ankle 205 of the leg orthosis 107 are driven according to the torque command to assist the user 108 in walking. However, the sampling time of the torque command is adjusted appropriately by the controller 106.

  In the sections [t3, t4], [t7, t8], [t9, t10], [t13, t14] immediately after the movement intention is switched, the non-healthy leg knee angle and the non-healthy leg ankle angle are fed back as described above. The non-healthy leg knee angle and the non-healthy leg ankle angle are controlled by the feedforward control in other sections.

  As described above, when the user 108 wearing the leg brace 107 is about to fall down, the walking control is performed so as to keep the horizontal movement speed and the trunk vertical position within a desired range by feedback control. Walking assistance is performed so that the unhealthy knee angle and the upper unhealthy leg ankle angle are operated as intended by the feedforward control.

  According to the present embodiment, the motion intention of the user 108 is estimated based only on the state detection value of the unhealthy leg thigh angle, and the motion of the unhealthy leg knee and the unhealthy leg ankle can be assisted. The leg brace 107 can be realized without the need for an expensive detector on the trunk 201 or the healthy leg 207, with reduced wiring, small size, light weight, and low price.

  In the Fourier coefficient storage unit 101, it is determined that the Fourier coefficient stored values having an error in the set range are the same walking pattern, and only the average value is stored, so that the memory usage can be reduced, and the leg brace. 107 can be constructed at low cost.

  In addition, the feedback control of the non-healthy leg knee angle and the non-healthy leg ankle angle is performed according to any control law such as position P speed PI control, position P speed IP control, position PID control, state feedback control, sliding mode control. Even if it uses, it is realizable similarly.

  In the present embodiment, it is assumed that walking assistance is performed when the user 108 cannot move freely below the knee of one leg, but the present invention can be similarly applied to various movement assistance such as movement assistance of an arm.

  As described above, the present embodiment can estimate the operation intention of the user 108 based on only the state detection values from a small number of detectors, so that the leg brace 107 is easy to wear with reduced wiring, small size, light weight and low cost. Thus, the operation of the user 108 can be assisted.

  The motion assist device of this embodiment can assist the user's motion by estimating the motion intention of the user with high accuracy using only a few inexpensive detectors. This motion assist device can be widely applied mainly in the medical field such as prosthetic hand, prosthetic leg, elderly motion assist device, limb rehabilitation device of patients who have paralyzed part of their limbs, or patients who have partially cut their limbs .

  Specific examples of the present invention have been described in detail above, but these are merely examples and do not limit the scope of the claims. The technology described in the claims includes various modifications and changes of the specific examples illustrated above.

  The technical elements described in this specification or the drawings exhibit technical usefulness alone or in various combinations, and are not limited to the combinations described in the claims at the time of filing. In addition, the technology exemplified in this specification or the drawings can achieve a plurality of objects at the same time, and has technical usefulness by achieving one of the objects.

10: Movement assisting device 12L: Leg orthosis 14L: Upper link 16L: Lower link 18L: Sole link 20aL: Waist joint 20bL: Knee joint 20cL: Ankle joint 21: Encoder (leg sensor)
30: Support link 32: Motor 40, 100: Controller 101: Fourier coefficient memory 102: Fourier coefficient calculator 103: Motion intention estimator 104: Target motion generator 105: Each joint command generator 106: Controller 107: Leg Equipment 109: State detector 110: Sample time adjuster

Claims (6)

It is a movement assist device that assists human movement,
An actuator that applies torque to a specific joint of a person,
A sensor that measures the movement pattern of a person's default site;
The first motion pattern group describing different motion patterns of the predetermined part is stored, and the target angle pattern for realizing each of the different second motion pattern groups of the specific joint is set as the first motion pattern group. A storage unit storing the information in association with each of the
A controller that controls the actuator;
The controller is equipped with
A first operation pattern closest to the measured operation pattern is selected from the first operation pattern group;
An operation assisting device that controls an actuator so that an angle of a specific joint follows a target angle pattern associated with a selected first operation pattern. The storage unit stores a first Fourier coefficient set obtained by Fourier transforming the time series data of each first operation pattern,
The controller calculates a second Fourier coefficient set obtained by performing Fourier transform on the time series data of the measured motion pattern, compares the second Fourier coefficient set with each first Fourier coefficient set, and calculates the second Fourier coefficient set. The motion assisting device according to claim 1, wherein the first motion pattern corresponding to the first Fourier coefficient set having the smallest difference from the first Fourier coefficient set is selected.   3. The motion assisting device according to claim 2, wherein the controller selects the first motion pattern corresponding to the first Fourier coefficient set having the smallest mean square error with respect to the second Fourier coefficient set. 4.   The controller stores the measured operation pattern in the storage unit in association with the selected first operation pattern, and weights the past measured operation pattern group stored in association with the first operation pattern. The motion assisting device according to claim 2, wherein the first motion pattern stored by averaging is updated.   5. The motion assisting apparatus according to claim 4, wherein the weighted average weight is a forgetting factor that decreases as the elapsed time from the stored time increases.   6. The controller according to claim 2, wherein the controller shortens a sampling time when the sensor measures an operation pattern of the predetermined part when the speed of the predetermined part detected by the sensor is larger than a threshold value. The motion auxiliary device according to Item 1.

Images