CN110977961A - Motion information acquisition system of self-adaptive power-assisted exoskeleton robot - Google Patents
Motion information acquisition system of self-adaptive power-assisted exoskeleton robot Download PDFInfo
- Publication number
- CN110977961A CN110977961A CN201911084056.9A CN201911084056A CN110977961A CN 110977961 A CN110977961 A CN 110977961A CN 201911084056 A CN201911084056 A CN 201911084056A CN 110977961 A CN110977961 A CN 110977961A
- Authority
- CN
- China
- Prior art keywords
- sensor
- data
- motion
- connecting rod
- exoskeleton robot
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1628—Programme controls characterised by the control loop
- B25J9/1633—Programme controls characterised by the control loop compliant, force, torque control, e.g. combined with position control
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/0006—Exoskeletons, i.e. resembling a human figure
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
- B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
- B25J9/00—Programme-controlled manipulators
- B25J9/16—Programme controls
- B25J9/1656—Programme controls characterised by programming, planning systems for manipulators
- B25J9/1664—Programme controls characterised by programming, planning systems for manipulators characterised by motion, path, trajectory planning
Landscapes
- Engineering & Computer Science (AREA)
- Robotics (AREA)
- Mechanical Engineering (AREA)
- Rehabilitation Tools (AREA)
- Manipulator (AREA)
Abstract
The invention discloses a motion information acquisition system of a self-adaptive power-assisted exoskeleton robot. The system comprises nine sensors attached to a big arm connecting rod, a small arm connecting rod, a back connecting rod, a thigh connecting rod, a shank connecting rod and an ankle joint connecting rod of the exoskeleton robot, and a sole sensor arranged in an exoskeleton robot shoe; the nine-axis sensor is used for measuring pitch, roll and yaw angle information when each connecting rod moves and transmitting the information to the computer end; the sole sensor collects sole pressure information and transmits the information to the computer terminal. The system can accurately acquire joint angle information and speed information in real time for identifying the walking process of a wearer wearing the exoskeleton robot, identify the motion state and acquire different personalized simple assistance exoskeleton motion acquisition configuration schemes.
Description
Technical Field
The invention relates to the field of intelligent robots, in particular to a motion information acquisition system of a self-adaptive power-assisted exoskeleton robot.
Background
As the assistance application, the exoskeleton can enhance the limb capacity of a wearer or directly bear external load, so that the energy consumption of a human body is effectively reduced, and the exoskeleton can be used as an assistance instrument for long-distance load walking for military use or completing the action of carrying objects. The exoskeleton can be applied to military affairs, medical treatment, post rescue and even daily life in the future.
The power-assisted exoskeleton robot comprises a load moving type exoskeleton robot for bearing a heavy object and an operation enhanced exoskeleton robot for moving and lifting the heavy object, and the load moving type exoskeleton robot comprises a rigid and flexible power-assisted exoskeleton robot, and the driving modes of the rigid and flexible power-assisted exoskeleton robot mainly comprise hydraulic driving and motor driving.
In the power-assisted exoskeleton robot, a person is a control center, an exoskeleton identifies and predicts the motion state of the human body according to motion information of the person, and then calculates the motion torque of each joint, so that a hydraulic cylinder is controlled to drive the exoskeleton joints to move, the exoskeleton is ensured to rapidly and accurately power the human body to move, and therefore, the acquisition of the human body motion information through a plurality of sensors is the basis for realizing the power-assisted motion of the exoskeleton.
For the configuration of multiple sensors, in the existing exoskeleton human body information acquisition system, although a large number of sensors are preferred by most researchers, in the actual exoskeleton using process, too many sensors are placed, which may cause redundancy and easily generate errors, and meanwhile, the complexity of the classification problem is increased, and too few sensors are placed, which greatly affects the improvement of the recognition rate. Therefore, it is very necessary for the power-assisted exoskeleton robot to select appropriate sensor information in different work tasks.
Disclosure of Invention
The invention mainly solves the technical problem of providing a motion information acquisition system of a self-adaptive power-assisted exoskeleton robot, and solves the problems that in the prior art, the motion state information data of the exoskeleton robot is difficult to effectively acquire, and the configuration of a sensor is reasonably and optimally selected by utilizing the motion state information data.
In order to solve the technical problems, the invention adopts a technical scheme that a motion information acquisition system of a self-adaptive power-assisted exoskeleton robot is provided, which comprises nine sensors attached to a big arm connecting rod, a small arm connecting rod, a back connecting rod, a thigh connecting rod, a shank connecting rod and an ankle joint connecting rod of the exoskeleton robot and a sole sensor arranged in an exoskeleton robot shoe; the nine-axis sensor is used for measuring pitch, roll and yaw angle information when each connecting rod moves and transmitting the information to the computer end; the sole sensor collects sole pressure information and transmits the information to the computer terminal.
In another embodiment of the motion information acquisition system of the self-adaptive power-assisted exoskeleton robot, the nine-axis sensor is a nine-axis inertial sensor based on an MEMS, integrates a high-precision gyroscope, an accelerometer and a geomagnetic field sensor, and adopts a microprocessor and a dynamic solution and Kalman dynamic filter algorithm to precisely measure the angular velocity and acceleration information of human motion.
In another embodiment of the adaptive power-assisted exoskeleton robot motion information acquisition system, the sole sensor comprises a plurality of single-point type flexible thin film pressure sensor pieces and a pressure signal processing module, and the pressure signal processing module controls the sampling work of each pressure sensor piece and calculates the pressure of each pressure sensor piece and the pressure center of the whole foot.
In another embodiment of the adaptive power-assisted exoskeleton robot motion information acquisition system of the present invention, the signal acquisition method of the nine-axis sensor and the sole sensor comprises the steps of:
the equipment calibration, namely, the zero offset of the accelerometer is removed through the accelerometer calibration on the nine-axis sensor, and the magnetic field calibration is used for removing the zero offset of the magnetic field sensor; sampling the sole sensors of the left foot and the right foot for multiple times to obtain zero offset values of the pressure sensor pieces, and eliminating the zero offset values of the pressure sensor pieces in the following walking process;
in the data acquisition process, the pressure sensor sheets of the plantar pressure sensor are sampled under the control of the pressure signal processing module, the voltage value corresponding to each pressure sensor sheet is obtained through low-pass filtering processing, and corresponding plantar pressure data are obtained through data conversion; respectively sampling by an accelerometer, a gyroscope and a geomagnetic field sensor, and solving pitching, rolling and yaw angles of outer skeleton thighs and shanks in the walking process through Kalman filtering;
and data storage, namely constructing a human body assistance exoskeleton motion information database, filtering, amplifying, denoising and discretizing collected gait data of a plurality of periods, and then storing the gait data to obtain the gait data at any moment in each period, wherein the gait data comprises joint angle data, lower limb pitching, rolling, yaw angle and sole pressure data.
In another embodiment of the adaptive power-assisted exoskeleton robot motion information acquisition system of the present invention, the data conversion corresponds to a conversion method comprising:
f(p)＝Ag(V,K1,K2)3+Bg(V,K1,K2)2+Cg(V,K1,K2)+D，
wherein A, B, C, D, K1,K2And determining a value for a system configuration file according to the model of the sensor, wherein V is a voltage value obtained by collection.
In another embodiment of the self-adaptive power-assisted exoskeleton robot motion information acquisition system, in the data acquisition process, a computer end is cascaded with each sensor through an MODBUS protocol, sends a request to each sensor, calls data in the sensor and returns a data value; the data is returned to the computer end through the 485 serial port, and the information is synchronously stored into the corresponding file name through the read-write thread.
In another embodiment of the adaptive power-assisted exoskeleton robot motion information acquisition system, based on acquired data, the system further comprises the steps of identifying a motion mode, identifying the motion mode by using a support vector machine algorithm, and inputting data of the support vector machine algorithm, wherein the data comprises inertial motion data acquired by a nine-axis sensor based on an MEMS and sole pressure data acquired by a sole sensor.
In another embodiment of the adaptive power-assisted exoskeleton robot motion information acquisition system, the system further comprises 7 motion modes of classifying motion modes, including walking on flat ground, ascending stairs, descending stairs, ascending slopes, descending slopes, lifting and placing.
In another embodiment of the motion information acquisition system of the adaptive power-assisted exoskeleton robot, a nonlinear model is used for classifying the 7 motion modes, a support vector machine classifier is constructed by adopting a one-to-one method, and then a training set T is input:
T＝{(x1,y1),(x2,y2),...,(xn,yn)},xi∈Rd,Rdrepresenting a positive real number set, yi∈{-1，+1},i＝1,2,...,n,xiInput feature vectors, y, for corresponding trainingiFor the labels corresponding to the training output vectors, the classifier is:
w and b are parameters of hyperplane, C is more than 0 and is a penalty parameter, which represents the penalty degree of the misclassification sample, and the relaxation variable ξ can be controlled by adjusting the size of CiThe invalid data points are effectively removed, and the overfitting phenomenon is avoided;
based on the formula, by introducing a Lanrange multiplier aiThe final discriminant function is derived as:
in another embodiment of the adaptive power-assisted exoskeleton robot motion information acquisition system, the system further comprises a classifier-based sensor configuration optimization method, and the method comprises the following steps:
in a first step, the excess sensors are configured to produce an excess of different base classifiers, forming a classifier pool D ═ { D ═ D1,D2,...,DLIn which D isLFor the Lth base classifier, taking the classifier corresponding to each sensor as a base classifier of the identification system;
secondly, pre-training the appointed 7 motion modes and the appointed wearer by using a base classifier combination, collecting motion data of the appointed motion modes, and constructing an exoskeleton wearing training set;
thirdly, training by using base classifiers according to a basic exoskeleton wearing training set, and storing a training data subset C of each base classifier for error classification1,C2,...,CmM is the number of the base classifiers, then the exoskeleton worn by the wearer is tested, and the degree of the experimental action of the assisting exoskeleton worn by the wearer belonging to one of the wrongly-classified data subsets is measured by using a support vector machine method; is provided with Ci＝{ni1,ni2,...,nimIs the data subset misclassified by the ith generated base classifier, where nijFor the j-th data in the data subset, let the motion state set Z ═ x1,x2,...,xn}，xiRepresenting the ith motion state in the set of motion states Z.
A fourth step of measuring said set of motion states Z and each misclassified training data subset CiEach instance n inijDistance d ofij：
Wherein x iskThe k-th motion state, x, of the set of motion states ZijkFor the training data subset C that is misclassifiediEach instance n inijIn the k-th motion state, if the two are the same, f (x)k,xijk) Is 0, otherwise f (x)k,xijk) Is 1;
the fifth step, measuring the motion state set Z and each of the misclassified training data subsets CiDistance d ofi：
And sixthly, measuring the possibility of correctly classifying the motion state by each base classifier according to the value taking condition of the test example, then dynamically distributing the weight to the base classifiers, and after the weight is optimized according to the continuous change of the weight, dividing the sensor-level classifier into a selection part and a temporary forgetting part, and feeding back the selection part and the temporary forgetting part to the sensor end to realize the on-off of the sensor.
The invention has the beneficial effects that: the invention discloses a motion information acquisition system of a self-adaptive power-assisted exoskeleton robot. The system comprises nine sensors attached to a big arm connecting rod, a small arm connecting rod, a back connecting rod, a thigh connecting rod, a shank connecting rod and an ankle joint connecting rod of the exoskeleton robot, and a sole sensor arranged in an exoskeleton robot shoe; the nine-axis sensor is used for measuring pitch, roll and yaw angle information when each connecting rod moves and transmitting the information to the computer end; the sole sensor collects sole pressure information and transmits the information to the computer terminal. The system can accurately acquire joint angle information and speed information in real time for identifying the walking process of a wearer wearing the exoskeleton robot, identify the motion state and acquire different personalized simple assistance exoskeleton motion acquisition configuration schemes.
Drawings
Fig. 1 is a schematic diagram of distribution of nine-axis sensors in a human body according to an embodiment of a motion information collection system of an adaptive power-assisted exoskeleton robot of the present invention;
FIG. 2 is a schematic view of a single-point flexible thin film pressure sensor patch distributed on a sole of a foot according to another embodiment of the adaptive power-assisted exoskeleton robot motion information acquisition system;
FIG. 3 is a schematic diagram of the components and operation of another embodiment of a motion information collection system for an adaptive power assist exoskeleton robot according to the present invention;
fig. 4 is a data collection flow chart of another embodiment of the adaptive power assistance exoskeleton robot motion information collection system according to the invention.
Detailed Description
In order to facilitate an understanding of the invention, the invention is described in more detail below with reference to the accompanying drawings and specific examples. Preferred embodiments of the present invention are shown in the drawings. This invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.
It is to be noted that, unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.
The invention discloses an adaptive power-assisted exoskeleton robot motion information acquisition system, which comprises nine-axis sensors and a sole sensor, wherein the nine-axis sensors are attached to a large arm connecting rod, a small arm connecting rod, a back connecting rod, a thigh connecting rod, a shank connecting rod and an ankle joint connecting rod of an exoskeleton robot; the nine-axis sensor is used for measuring pitch, roll and yaw angle information when each connecting rod moves and transmitting the information to the computer end; the sole sensor collects sole pressure information and transmits the information to the computer terminal.
Preferably, the nine-axis sensor is a nine-axis inertial sensor based on MEMS, integrates a high-precision gyroscope, an accelerometer and a geomagnetic field sensor, and adopts a microprocessor and a dynamic calculation and Kalman dynamic filtering algorithm to precisely measure the angular velocity and acceleration information of the human motion. Preferably, as shown in fig. 1, there are 11 nine-axis sensors respectively distributed on the back n0, the right upper arm n1, the right lower arm n2, the left lower arm n3, the left upper arm n4, the right thigh n5, the left thigh n6, the right calf n7, the left calf n8, the right ankle n9 and the left ankle n 10.
Preferably, the sole sensor comprises a plurality of single-point type flexible thin film pressure sensor pieces and a pressure signal processing module, as shown in fig. 2, 16 single-point type flexible thin film pressure sensor pieces are arranged, wherein 9 sensor pieces with the serial numbers of 3 to 11 are mainly distributed in a sole area, 4 sensor pieces with the serial numbers of 12 to 15 are distributed in a heel area, and two sensor pieces with the serial numbers of 1 and 2 are distributed in a middle outer area of the sole, so that sole pressures of the sole, the heel and the outer side of the sole can be fully collected, and a sole pressure cloud chart is constructed. The pressure signal processing module controls the sampling work of each pressure sensor piece and calculates the pressure of each pressure sensor piece.
Preferably, the single-point flexible film pressure sensor sheet and the pressure signal processing module are arranged in a sensor insole, and the insole is connected with the pressure signal processing module through a silica gel flexible lead. The signal processing module outputs pressure information of digital signals, a USB interface is connected to a computer end, and the computer end can display pressure distribution conditions in real time and can also store pressure information data.
Preferably, as shown in fig. 3, the nine-axis sensor and the sole sensor are interoperated with the computer terminal through a plurality of sub-threads, respectively. Further, the method comprises 1 sole sensor acquisition sub-thread, namely 1 thread, 4 nine-axis sensor acquisition sub-threads, namely 2 to 5 threads, and 1 read-write data sub-thread, namely 6 thread. Setting an identifier in a main thread, sending acquisition instructions in a multi-thread concurrent manner, activating when the main thread receives an acquisition ending signal, sending a new acquisition control signal, and starting a new acquisition task after the sub-thread receives the signal.
Further preferably, in combination with fig. 3, fig. 4 shows that the signal acquisition method of the nine-axis sensor and the sole sensor includes:
the method comprises the following steps that S101, equipment calibration is carried out, zero offset of an accelerometer is removed through meter adding calibration on a nine-axis sensor, and magnetic field calibration is used for removing the zero offset of a magnetic field sensor; sampling the sole sensors of the left foot and the right foot for multiple times to obtain zero offset values of the pressure sensor pieces, and eliminating the zero offset values of the pressure sensor pieces in the following walking process;
and S102, acquiring data, wherein in the data acquisition process, the pressure sensor sheets of the sole sensor are sampled under the control of the pressure signal processing module, the voltage value corresponding to each pressure sensor sheet is obtained through low-pass filtering processing, corresponding sole pressure data are obtained through data conversion, and the conversion method of converting voltage into sole pressure comprises the following steps:
f(p)＝Ag(V,K1,K2)3+Bg(V,K1,K2)2+Cg(V,K1,K2)+D，
wherein A, B, C, D, K1,K2And determining a value for a system configuration file according to the model of the sensor, wherein V is a voltage value obtained by collection.
Preferably, in the data acquisition process, the computer end is cascaded with the sensor through an MODBUS protocol, sends a request to the sensor, calls data in the sensor and returns a data value, and the acceleration and angle change measured by the sensor in the motion process are obtained through calculation. And respectively sampling by an accelerometer, a gyroscope and a geomagnetic sensor, and solving the pitching, rolling and yaw angles of the thighs and the shanks of the exoskeleton in the walking process through Kalman filtering.
Preferably, in the data acquisition process, data are transmitted back to the computer terminal through the 485 serial port, and information is synchronously stored into the TXT document corresponding to the file name through the read-write thread, so that the gait analysis and the predicted data processing in the next step are facilitated.
And S103, storing data, constructing a human body assistance exoskeleton motion information database, filtering, amplifying, denoising and discretizing collected gait data of multiple periods, obtaining gait data at any moment in a period, and storing the gait data, wherein the gait data comprises joint angle data, lower limb pitching, rolling, yaw angle and sole pressure data.
Preferably, a plurality of groups of data are repeatedly acquired for a preset number of sample groups and are subjected to data processing, and a database is established. Further, during data processing, comparing the difference between different periods of the same sample and the difference between different samples, integrating data and optimizing, removing data with larger difference, removing data with the proportion deviating from the overall average value exceeding a preset threshold value, such as exceeding 20%, and averaging the remaining data to obtain sample data in the database.
Furthermore, based on the system and the acquired data, the motion mode identification is further included, because to control the loaded exoskeleton, the motion mode of the exoskeleton must be known so as to take the action corresponding to the corresponding mode for assisting, and the fast and accurate identification of the motion mode makes the system more intelligent.
Preferably, the motion pattern recognition is performed by using a support vector machine algorithm (SVM), and the input signal of the SVM comprises inertial motion data acquired by a MEMS-based nine-axis sensor and plantar pressure data acquired by a plantar pressure sensor.
For the signals acquired, features are extracted by using a Principal Component Analysis (PCA), and after the features are acquired, the sensor information is normalized, and the size of the information is normalized to a [0,1] interval.
Preferably, the motion modes are classified, and the motion modes mainly comprise 7 motion modes of flat walking, ascending stairs, descending stairs, ascending slopes, descending slopes, lifting and placing. Nine-axis inertial sensors based on MEMS are installed at the upper arm, the lower arm, the back, the thigh, the crus and the ankle and are used for measuring the pitching, rolling and yaw angles of the upper arm, the lower arm, the thigh, the crus and the trunk of a human body in the walking process.
In the motion pattern recognition process, a nonlinear model is used for specific 7 motion pattern classifications, a support vector machine classifier is built by adopting a one-to-one method, and then a training set is input:
T＝{(x1,y1),(x2,y2),...,(xn,yn)}，xi∈Rd,Rdrepresenting a positive real number set, yiE { -1, +1}, i ═ 1,2, ·, n, where x isiInputting feature vectors for corresponding training, including normalized inertial sensor data and plantar pressure sensor data, yiDetermining for the labels corresponding to the training output vectors which mode the training input vectors are, the classifier being:
w and b are parameters of hyperplane, C is more than 0 and is a penalty parameter, which represents the penalty degree of the misclassification sample, and the relaxation variable ξ can be controlled by adjusting the size of CiThe invalid data points are effectively removed, and the overfitting phenomenon is avoided; in addition, on the basis of the formula, by introducing a Lanrange multiplier aiThe final discriminant function is derived as:
by using the algorithm to train 252 one-to-one support vector machine classification models, each sensor corresponds to 21 one-to-one support vector machine classification models, which are 7 motion patterns, if the models are separated one by one, 21 support vector machines (6+1) × 6/2 ═ 12 are required, 12 are 12 sensors, 11 inertial sensors and 1 plantar pressure sensor. The support vector machine has the core idea that through establishing an optimal classification surface, the positive class and the negative class can be isolated maximally in a feature space, and in order to find an optimal segmentation hyperplane, each classifier divides the feature space into two parts, one part is the positive class and the other part is the negative class, so that a data set is classified, and the motion mode of input data can be judged.
The collected data can be trained offline, the trained model is imported into online motion pattern identification, and the identification result is post-processed by using a majority vote Method (MVA), so that the identification precision is improved, and configuration optimization selection is performed on the sensor.
Further preferably, the method further comprises a sensor configuration optimization method based on the classifier, and mainly comprises the following steps:
in a first step, the excess sensors are configured to produce an excess of different base classifiers, forming a classifier pool D ═ { D ═ D1,D2,...,DLIn which D isLFor the Lth base classifier, taking the classifier corresponding to each sensor as a base classifier of the identification system;
secondly, pre-training the appointed 7 motion modes and the appointed wearer by using a base classifier combination, collecting motion data of the appointed motion modes, and constructing an exoskeleton wearing training set;
thirdly, training by using base classifiers according to a basic exoskeleton wearing training set, and storing a training data subset C of each base classifier for error classification1,C2,...,Cm(there are m basis classifiers). And then the exoskeleton worn by the wearer is tested, and the degree of the experimental action of the assisting exoskeleton worn by the wearer belonging to a certain wrongly-divided data subset is measured by using a support vector machine method.
Preferably, let Ci＝{ni1,ni2,...,nimIs the data subset misclassified by the ith generated base classifier, where nijIs given to the jth data in the subset. Let the motion state set Z ═ x1,x2,...,xn}，xiRepresenting the ith motion state in the set of motion states Z.
The fourth step is to measure the set of motion states Z made and each of the misclassified training data subsets CiEach instance n inijThe distance of (c):
wherein x iskThe k-th motion state, x, of the set of motion states ZijkFor the training data subset C that is misclassifiediEach instance n inijIn the k-th motion state, if the two are the same, f (x)k,xijk) Is 0, otherwise f (x)k,xijk) Is 1;
the fifth step, measuring the motion state set Z and each of the misclassified training data subsets CiDistance d ofi：
Distance diIndicating the extent to which the test action belongs to the training data subset and indirectly how likely it can be correctly classified by the base classifier that wrongly classifies the training subset, diA large value indicates that the test motion state does not belong to this misclassified subset to a lesser extent, i.e. is less likely to be paired by the corresponding base classifier, and should therefore be given a lower weight,
preferably, 1/d will be used in the dynamic adjustment of the base classifier weightsiThe base classifier weights are dynamically adjusted on a basis and normalized to a sum of 1, and then normalized 1/d is usediAs the weight of the i +1 th base classifier, the weight of the i-th base classifier is:
and sixthly, measuring the possibility of correctly classifying the motion state by each base classifier according to the value taking condition of the test example, then dynamically distributing the weight to the base classifiers, and after the weight is optimized according to the continuous change of the weight, dividing the sensor-level classifier into a selection part and a temporary forgetting part, and feeding back the selection part and the temporary forgetting part to the sensor end to realize the on-off of the sensor.
The weight of the time-based classifier can be changed according to the change of the test motion state every time classification is carried out, and according to the continuous change of the weight, after the weight is continuously optimized, the sensor-level classifier can be divided into a selection part and a temporary forgetting part and fed back to the sensor end to realize the opening and closing of the sensor, so that the effects of saving energy and optimizing the result are achieved.
The invention aims to provide a motion information acquisition system and a motion information acquisition method for a self-adaptive power-assisted exoskeleton robot, which can accurately acquire joint angle information and speed information for identifying the walking process of a wearer wearing the exoskeleton robot in real time through an accelerometer and a gyroscope, identify a motion state, and adaptively turn on or turn off a plurality of sensors in real time to acquire personalized and simplified power-assisted exoskeleton motion acquisition configuration schemes for different operation tasks (such as load advancing and weight hanging), different acquisition scenes (mountainous regions and ship-borne people) and different wearers (different heights and different weights). At different operation tasks, gather the scene, in the use of different wearers promptly, required sensor quantity and type diverse, if at the heavy burden uphill in-process, required sensor obtains low limbs motion angle for low limbs sensor to and plantar pressure sensor obtains the focus removal condition of walking in-process, do not need the participation of upper limbs and partial truck sensor, and at the in-process of carry the heavy object, need the upper limbs sensor to discern, judge the motion condition of upper limbs, the low limbs sensor need not participate in. When the system runs for a period of time, sensors which are frequently forgotten can be closed to collect, and personalized self-adaptive optimization of sensor configuration is achieved.
Therefore, the invention discloses a motion information acquisition system of a self-adaptive power-assisted exoskeleton robot. The system comprises nine sensors attached to a big arm connecting rod, a small arm connecting rod, a back connecting rod, a thigh connecting rod, a shank connecting rod and an ankle joint connecting rod of the exoskeleton robot, and a sole sensor arranged in an exoskeleton robot shoe; the nine-axis sensor is used for measuring pitch, roll and yaw angle information when each connecting rod moves and transmitting the information to the computer end; the sole sensor collects sole pressure information and transmits the information to the computer terminal. The system can accurately acquire joint angle information and speed information in real time for identifying the walking process of a wearer wearing the exoskeleton robot, identify the motion state and acquire different personalized simple assistance exoskeleton motion acquisition configuration schemes.
The above description is only an embodiment of the present invention, and not intended to limit the scope of the present invention, and all equivalent structural changes made by using the contents of the present specification and the drawings, or applied directly or indirectly to other related technical fields, are included in the scope of the present invention.
Claims (10)
1. A self-adaptive power-assisted exoskeleton robot motion information acquisition system is characterized by comprising nine sensors attached to a large arm connecting rod, a small arm connecting rod, a back connecting rod, a thigh connecting rod, a shank connecting rod and an ankle joint connecting rod of an exoskeleton robot and a sole sensor arranged in an exoskeleton robot shoe; the nine-axis sensor is used for measuring pitch, roll and yaw angle information when each connecting rod moves and transmitting the information to the computer end; the sole sensor collects sole pressure information and transmits the information to the computer terminal.
2. The adaptive power-assisted exoskeleton robot motion information acquisition system of claim 1, wherein the nine-axis sensor is a nine-axis MEMS-based inertial sensor, integrates a high-precision gyroscope, an accelerometer and a geomagnetic field sensor, and adopts a microprocessor and a dynamic solution and Kalman dynamic filter algorithm to precisely measure angular velocity and acceleration information of human motion.
3. The system for acquiring motion information of an adaptive power-assisted exoskeleton robot of claim 2, wherein the sole sensor comprises a plurality of single-point flexible thin film pressure sensor plates and a pressure signal processing module, and the pressure signal processing module controls sampling of each pressure sensor plate and calculates the pressure of each pressure sensor plate and the pressure center of the whole foot.
4. The adaptive power-assisted exoskeleton robot motion information acquisition system of claim 3, wherein the signal acquisition method of the nine-axis sensor and the sole sensor comprises the steps of:
the equipment calibration, namely, the zero offset of the accelerometer is removed through the accelerometer calibration on the nine-axis sensor, and the magnetic field calibration is used for removing the zero offset of the magnetic field sensor; sampling the sole sensors of the left foot and the right foot for multiple times to obtain zero offset values of the pressure sensor pieces, and eliminating the zero offset values of the pressure sensor pieces in the following walking process;
in the data acquisition process, the pressure sensor sheets of the plantar pressure sensor are sampled under the control of the pressure signal processing module, the voltage value corresponding to each pressure sensor sheet is obtained through low-pass filtering processing, and corresponding plantar pressure data are obtained through data conversion; respectively sampling by an accelerometer, a gyroscope and a geomagnetic field sensor, and solving pitching, rolling and yaw angles of outer skeleton thighs and shanks in the walking process through Kalman filtering;
and data storage, namely constructing a human body assistance exoskeleton motion information database, filtering, amplifying, denoising and discretizing collected gait data of a plurality of periods, and then storing the gait data to obtain the gait data at any moment in each period, wherein the gait data comprises joint angle data, lower limb pitching, rolling, yaw angle and sole pressure data.
5. The adaptive power-assisted exoskeleton robot motion information collection system of claim 4, wherein the data transformation corresponds to a transformation method that is:
f(p)＝Ag(V,K1,K2)3+Bg(V,K1,K2)2+Cg(V,K1,K2)+D，
wherein A, B, C, D, K1,K2And determining a value for a system configuration file according to the model of the sensor, wherein V is a voltage value obtained by collection.
6. The adaptive power-assisted exoskeleton robot motion information acquisition system of claim 5, wherein in the data acquisition process, the computer end is cascaded with each sensor through MODBUS protocol, sends requests to each sensor, calls data in the sensor and returns data values; the data is returned to the computer end through the 485 serial port, and the information is synchronously stored into the corresponding file name through the read-write thread.
7. The adaptive power-assisted exoskeleton robot motion information collection system of claim 6, wherein based on the data obtained by collection, further comprising identification of motion patterns, identification of motion patterns using support vector machine algorithm, the data input to the support vector machine algorithm comprises inertial motion data collected by the MEMS-based nine-axis sensors and sole pressure data collected by the sole sensors.
8. The adaptive power assisted exoskeleton robot motion information collection system of claim 7 further comprising classifying the 7 motion patterns including walking on level ground, ascending stairs, descending stairs, ascending slopes, descending slopes, lifting and placing.
9. The adaptive power-assisted exoskeleton robot motion information collection system of claim 8, wherein a nonlinear model is used for the 7 motion mode classification, a one-to-one method is adopted to construct a support vector machine classifier, and then a training set T:
T＝{(x1,y1),(x2,y2),...,(xn,yn)},xi∈Rd,Rdrepresenting a positive real number set, yi∈{-1,+1},i＝1,2,...,n,xiInput feature vectors, y, for corresponding trainingiFor the labels corresponding to the training output vectors, the classifier is:
w and b are parameters of hyperplane, C is more than 0 and is a penalty parameter, which represents the penalty degree of the misclassification sample, and the relaxation variable ξ can be controlled by adjusting the size of CiThe invalid data points are effectively removed, and the overfitting phenomenon is avoided;
based on the formula, by introducing a Lanrange multiplier aiThe final discriminant function is derived as:
10. the adaptive power assist exoskeleton robot motion information collection system of claim 9 further comprising a classifier-based sensor configuration optimization method comprising the steps of:
in a first step, an excess of sensors is configured to produce an excess of different base classifiers, forming a classifier pool D ═ D { (D)1,D2,...,DLIn which D isLFor the Lth base classifier, taking the classifier corresponding to each sensor as a base classifier of the identification system;
secondly, pre-training the appointed 7 motion modes and the appointed wearer by using a base classifier combination, collecting motion data of the appointed motion modes, and constructing an exoskeleton wearing training set;
thirdly, training by using base classifiers according to a basic exoskeleton wearing training set, and storing a training data subset C of each base classifier for error classification1,C2,...,CmM is the number of the base classifiers, then the exoskeleton worn by the wearer is tested, and the degree of the experimental action of the assisting exoskeleton worn by the wearer belonging to one of the wrongly-classified data subsets is measured by using a support vector machine method; is provided with Ci＝{ni1,ni2,...,nimIs the data subset misclassified by the ith generated base classifier, where nijFor the j-th data in the data subset, let the motion state set Z ═ x1,x2,...,xn}，xiRepresenting the ith motion state in the set of motion states Z.
A fourth step of measuring said set of motion states Z and each misclassified training data subset CiEach instance n inijDistance d ofij：
Wherein x iskThe k-th motion state, x, of the set of motion states ZijkTo be misclassifiedTraining data subset CiEach instance n inijIn the k-th motion state, if the two are the same, f (x)k,xijk) Is 0, otherwise f (x)k,xijk) Is 1;
the fifth step, measuring the motion state set Z and each of the misclassified training data subsets CiDistance d ofi：
And sixthly, measuring the possibility of correctly classifying the motion state by each base classifier according to the value taking condition of the test example, then dynamically distributing the weight to the base classifiers, and after the weight is optimized according to the continuous change of the weight, dividing the sensor-level classifier into a selection part and a temporary forgetting part, and feeding back the selection part and the temporary forgetting part to the sensor end to realize the on-off of the sensor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911084056.9A CN110977961A (en) | 2019-11-07 | 2019-11-07 | Motion information acquisition system of self-adaptive power-assisted exoskeleton robot |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911084056.9A CN110977961A (en) | 2019-11-07 | 2019-11-07 | Motion information acquisition system of self-adaptive power-assisted exoskeleton robot |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110977961A true CN110977961A (en) | 2020-04-10 |
Family
ID=70083564
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911084056.9A Pending CN110977961A (en) | 2019-11-07 | 2019-11-07 | Motion information acquisition system of self-adaptive power-assisted exoskeleton robot |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110977961A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112536821A (en) * | 2020-12-02 | 2021-03-23 | 北方工业大学 | Exoskeleton design method for carrying load in situ |
| CN112641603A (en) * | 2020-12-17 | 2021-04-13 | 迈宝智能科技(苏州)有限公司 | Exoskeleton device and exoskeleton motion control method |
| CN112790951A (en) * | 2020-12-30 | 2021-05-14 | 上海傅利叶智能科技有限公司 | Gravity center recognition method, training assisting method and system and intelligent walking stick |
| CN113080879A (en) * | 2021-05-06 | 2021-07-09 | 陕西福音假肢有限责任公司 | Circuit and method for continuously and dynamically measuring joint movement angle |
| CN113442174A (en) * | 2021-05-27 | 2021-09-28 | 重庆理工大学 | Exoskeleton performance testing method, device and system |
| CN114332912A (en) * | 2021-11-22 | 2022-04-12 | 清华大学 | Human motion capture and joint stress analysis method based on IMU |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2660742A1 (en) * | 2012-05-02 | 2013-11-06 | Hocoma AG | Training apparatus |
| CN104224498A (en) * | 2014-09-24 | 2014-12-24 | 哈尔滨工业大学 | Exoskeleton robot system and kinematics extremity detection-based control method |
| KR101842465B1 (en) * | 2016-10-20 | 2018-03-27 | 국방과학연구소 | Lower limb exoskeleton robot and control method thereof |
| CN208926953U (en) * | 2018-02-13 | 2019-06-04 | 中山市沃倍特智能医疗机器人股份有限公司 | The ectoskeleton structure of intelligent exoskeleton robot |
| CN110141239A (en) * | 2019-05-30 | 2019-08-20 | 东北大学 | A kind of motion intention identification and installation method for lower limb exoskeleton |
-
2019
- 2019-11-07 CN CN201911084056.9A patent/CN110977961A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2660742A1 (en) * | 2012-05-02 | 2013-11-06 | Hocoma AG | Training apparatus |
| CN104224498A (en) * | 2014-09-24 | 2014-12-24 | 哈尔滨工业大学 | Exoskeleton robot system and kinematics extremity detection-based control method |
| KR101842465B1 (en) * | 2016-10-20 | 2018-03-27 | 국방과학연구소 | Lower limb exoskeleton robot and control method thereof |
| CN208926953U (en) * | 2018-02-13 | 2019-06-04 | 中山市沃倍特智能医疗机器人股份有限公司 | The ectoskeleton structure of intelligent exoskeleton robot |
| CN110141239A (en) * | 2019-05-30 | 2019-08-20 | 东北大学 | A kind of motion intention identification and installation method for lower limb exoskeleton |
Non-Patent Citations (5)
| Title |
|---|
| 付彬等: "Boosting算法中基分类器权重的动态赋值", 《广西师范大学学报：自然科学版》 * |
| 汪海燕等: "支持向量机理论及算法研究综述", 《计算机应用研究》 * |
| 罗莎等: "改进Adaboost算法的人体步态识别方法", 《控制工程》 * |
| 邓清龙: "下肢外骨骼机器人智能传感系统设计及应用", 《中国优秀硕士学位论文全文数据库：信息科技辑》 * |
| 陈小平等: "《无线传感器网络原理及应用》", 30 April 2017 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112536821A (en) * | 2020-12-02 | 2021-03-23 | 北方工业大学 | Exoskeleton design method for carrying load in situ |
| CN112536821B (en) * | 2020-12-02 | 2021-12-07 | 北方工业大学 | Exoskeleton design method for carrying load in situ |
| CN112641603A (en) * | 2020-12-17 | 2021-04-13 | 迈宝智能科技(苏州)有限公司 | Exoskeleton device and exoskeleton motion control method |
| CN112790951A (en) * | 2020-12-30 | 2021-05-14 | 上海傅利叶智能科技有限公司 | Gravity center recognition method, training assisting method and system and intelligent walking stick |
| CN113080879A (en) * | 2021-05-06 | 2021-07-09 | 陕西福音假肢有限责任公司 | Circuit and method for continuously and dynamically measuring joint movement angle |
| CN113442174A (en) * | 2021-05-27 | 2021-09-28 | 重庆理工大学 | Exoskeleton performance testing method, device and system |
| CN113442174B (en) * | 2021-05-27 | 2023-05-02 | 重庆理工大学 | Exoskeleton performance testing method, device and system |
| CN114332912A (en) * | 2021-11-22 | 2022-04-12 | 清华大学 | Human motion capture and joint stress analysis method based on IMU |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110977961A (en) | Motion information acquisition system of self-adaptive power-assisted exoskeleton robot | |
| CN110141239B (en) | Movement intention recognition and device method for lower limb exoskeleton | |
| CN112754468B (en) | Human body lower limb movement detection and identification method based on multi-source signals | |
| Hsu et al. | Human daily and sport activity recognition using a wearable inertial sensor network | |
| CN110303471B (en) | Power-assisted exoskeleton control system and control method | |
| CN110916679B (en) | Human body lower limb pose gait detection device and method | |
| US20210293643A1 (en) | Multi-Modal Fingertip Sensor With Proximity, Contact, And Force Localization Capabilities | |
| CN109940584A (en) | The detection method that a kind of exoskeleton robot and its detection human motion are intended to | |
| CN102799937A (en) | Lower limb movement track predication method under fusion of information of myoelectricity signal and joint angle | |
| Wang et al. | Human Gait Recognition System Based on Support Vector Machine Algorithm and Using Wearable Sensors. | |
| CN108334827B (en) | Gait identity authentication method based on intelligent shoe and intelligent shoe | |
| CN113043248A (en) | Transportation and assembly whole-body exoskeleton system based on multi-source sensor and control method | |
| CN103345626A (en) | Intelligent wheelchair static gesture identification method | |
| Khodabandelou et al. | A fuzzy convolutional attention-based GRU network for human activity recognition | |
| CN113314209A (en) | Human body intention identification method based on weighted KNN | |
| Song et al. | Adaptive neural fuzzy reasoning method for recognizing human movement gait phase | |
| Zheng et al. | Locomotion mode recognition with robotic transtibial prosthesis in inter-session and inter-day applications | |
| Hu et al. | A novel fusion strategy for locomotion activity recognition based on multimodal signals | |
| Sahoo et al. | Prediction of step length using neuro-fuzzy approach suitable for prosthesis control | |
| CN114224326A (en) | Wearable gait phase and action recognition device and method | |
| CN110751060B (en) | Portable motion mode real-time identification system based on multi-source signals | |
| CN110694252A (en) | Running posture detection method based on six-axis sensor | |
| CN116019443A (en) | Cardiopulmonary resuscitation chest compression compliance detection system and method | |
| Li et al. | A Novel Method of Pattern Recognition Based on TLSTM in lower limb exoskeleton in Many Terrains | |
| CN114831784A (en) | Lower limb prosthesis terrain recognition system and method based on multi-source signals |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200410 |