Hip power-assisted exoskeleton control method based on force position information fusion

Abstract

The invention relates to a hip power-assisted exoskeleton control method based on force and position information fusion, which realizes intelligent switching of multiple power-assisted modes by accurately identifying different motion states through multi-mode perception and fusion, and comprises the step of identifying different states through multi-source data such as a pitching angle and an angular velocity of an IMU (Inertial Measurement Unit) of the back of a body, an angle value of a hip motor encoder and the like. Corresponding control is carried out in different motion states; in the walking state, a zero-force dragging algorithm is adopted to reduce resistance, and torque compensation is carried out based on the hip joint speed to enhance the following performance; in the getting-up state, a power assisting algorithm combining PD control based on position errors and supporting force compensation is adopted; supporting force is provided in the squatting stop state and the getting-up stop state. The control part comprises a data processing module, a state recognition module, a control strategy module and a motor control module. The motion state can be accurately recognized, the power assisting efficiency is effectively improved, the physical consumption of a wearer is reduced, and the method is suitable for various task scenes.

Description

Technical Field

[0001] The present invention relates to the field of exoskeleton control technology, and specifically to a hip-assisted exoskeleton control method based on force-position information fusion, which can accurately identify human motion patterns, provide efficient and precise power-assisted control for the hip-assisted exoskeleton, and improve the coordinated motion performance of the exoskeleton and the human body. Background Art

[0002] With the continuous advancement of science and technology, exoskeleton robots have shown an increasingly broad application prospect in many fields such as logistics, industrial manufacturing, elderly care, and construction. It has an important significance in assisting human labor in carrying heavy objects and enhancing human movement ability. However, the current exoskeleton technology still faces many severe challenges.

[0003] Early power-assisted exoskeleton control was mainly controlled by a predetermined program. For example, the Israeli "exoskeleton" power-assisted device ReWalk, developed by Elger Medical Technology, uses a pair of crutches to help maintain body balance. It consists of an electric leg support, a body sensor, and a backpack. The backpack contains a computer control box and a rechargeable battery. The user can use a remote control belt to select a certain setting, such as standing, sitting, walking, climbing, etc., and then lean forward to activate the body sensor to put the mechanical leg in motion. It is mainly used to help paralyzed people regain the ability to walk. The power transmission adopts the method of motor-reducer-exoskeleton mechanism. The movement mode is mainly that the device drives the human body to move. The power size of the device is set by the control system and cannot be changed at any time according to the person's movement intention.

[0004] The military cooperation project of the University of California Berkeley - exoskeleton-assisted robot soldier suit. The device is called Berkeley Lower Extremity Exoskeleton or BLEEX, which was designed by the Advanced Defense Research Engineering Agency. It attempts to connect automatic mechanical struts to human legs to reduce the weight, so that infantry can travel longer distances with heavier loads. This set of equipment mainly consists of fuel supply and engine system, control and detection system, hydraulic transmission system and exoskeleton mechanism. People who use this device have to connect their legs to the legs of the mechanical exoskeleton through a transmission belt, and carry a large backpack equipped with an engine and control system on their backs. There is also space in the backpack to carry the payload. The power transmission process is: engine-hydraulic system-exoskeleton mechanism. The device can balance the weight of the device (50 kg), so that people can wear it without feeling the load, and the control system will ensure that its center of gravity is always on the user's feet. The backpack of the device can also carry 32 kg of weight. For the user, he only feels like he is carrying 2 kg. In addition to helping soldiers, the device could assist medical personnel in evacuating the wounded from danger zones or enable firefighters to climb more floors carrying heavy equipment.

[0005] Wearable robotic clothing will help troops improve their combat effectiveness and endurance. Raytheon Sarcos, the latest model developed by Raytheon, has claw-like hands. When American soldiers wear this equipment, their strength and endurance will be increased 20 times than normal. For Lockheed Martin's HULC model, the wearer can easily carry 200 pounds (about 91 kilograms) of objects and complete heavier tasks with less energy consumption. Currently, the U.S. Army Soldier Systems Center is conducting military trials on robotic exoskeletons.

[0006] Another military cooperation project in the United States, Raytheon Sarcos XOS, which represents the latest level of power-assisted exoskeleton robots, is the masterpiece of Dr. Steve Jacobsen, the mobile exoskeleton "XOS". The exoskeleton "XOS" was developed in secret after 7 years of research and development with a military research budget of 10 million US dollars provided by the Advanced Research Projects Agency (DARPA) of the United States Department of Defense in order to create superhuman soldiers. It represents the most advanced technology in the field of mechanical exoskeletons. Its control concept is the same as BLEEX. The control system determines the next action of the human body through the detection system and the microcomputer system, thereby determining how much power and speed to add to the human body, and also transmits the force to the exoskeleton mechanism through the hydraulic system, but it is a full-body armed exoskeleton, while BLEEX is a lower limb exoskeleton robot. The "XOS" movement is much more agile and powerful than the previous exoskeleton equipment. Using sensors attached to the body, it can respond to the body's movements without delay and output powerful force. When wearing "XOS", it can lift 90.7kg of weight while the human body feels only 9kg, and can lift 50-500 times continuously. But currently "XOS" has a major flaw, which is that its built-in battery can only be used for 40 minutes.

[0007] According to the above information, the above-mentioned military exoskeleton control technology is advanced, but because it is a military confidential technology, its control method has not been disclosed.

[0008] The Chinese invention patent with the authorization announcement number CN112949676B on July 8, 2022 discloses an adaptive motion mode recognition method for a flexible lower limb power-assisted exoskeleton robot, comprising the following steps: 1) obtaining a large amount of sensor data corresponding to the swinging dynamics in various motion modes as the motion information of the human body; 2) extracting motion features according to the motion information, and selecting the motion feature information corresponding to the minimum moment of the calf IMU; 3) normalizing the motion feature information and inputting it into the BP neural network for training, and the output result is the recognition probability of each motion mode, and the one with the largest recognition probability is the final motion mode recognition result; 4) obtaining the feature vector corresponding to the minimum moment of the calf IMU; 5) using the trained neural network model to perform motion mode recognition; 6) using the finite state transfer algorithm to correct the recognition result of the BP neural network. The present invention improves the motion mode by distinguishing the curve information position of the motion mode and performing motion mode recognition according to the feature information of the position.

[0009] The Chinese invention patent published on March 8, 2024 with publication number CN117653505A discloses a hip-assisted exoskeleton robot, a hip-assisted method, an apparatus and a medium, including: a waist belt, both left and right ends of the waist belt are provided with waist support members; two hip-joint drive modules, both hip-joint drive modules include a rotation drive assembly and a hip-joint connector, and both ends of the rotation drive assembly are respectively connected to the waist support member and the hip-joint connector; two leg-wearing assemblies, both leg-wearing assemblies include an upper leg rod and a lower leg rod, and both ends of the upper leg rod are respectively connected to the hip-joint connector and the lower leg rod, and the lower leg rod is provided with a leg belt; a control box is arranged at the front end of the waist belt, and the left and right ends of the control box are respectively connected to the waist support member through a chain, a controller is arranged in the control box, and the rotation drive assembly is electrically connected to the controller; the embodiment of the present application can be suitable for users of different body shapes, and can effectively provide assistance to the hip joints of the users. Summary of the invention

[0010] In view of the above problems existing in the prior art, a hip-assisted exoskeleton control method based on force-position information fusion is proposed. It can accurately identify the human body's motion pattern and provide efficient and accurate power-assisted control for the hip-assisted exoskeleton, thereby significantly improving its coordinated motion performance with the human body, effectively overcoming the shortcomings of the prior art, and greatly improving the performance and practicality of the power-assisted exoskeleton. The specific contents are as follows: The hip-assisted exoskeleton control method based on force-position information fusion of the present invention comprises the following steps: S1. Multimodal information collection and preprocessing The inertial sensor installed on the back is used to collect the body pitch angle and pitch angular velocity information, the absolute encoder is used to collect the hip joint angle information, and the torque sensor is used to collect the hip joint torque information. The collected inertial sensor data is filtered to remove noise interference, and the absolute encoder data is calibrated to ensure the accuracy of angle measurement. The torque sensor data is corrected for zero drift. S2. Motion state recognition and judgment The body pitch angle and angular velocity collected by the inertial sensor are combined with the angle value of the absolute encoder to make state machine switching judgments according to preset threshold conditions. The states include standing / walking state, squatting state, squatting stop state, standing up state, and standing up stop state. Set the body pitch angle threshold θ stand , body pitch angular velocity threshold v move , hip position difference threshold θ e , hip angular velocity product threshold v e The motion states include standing / walking state, squatting state, squatting stop state, standing up state, and standing up stop state. The specific judgment methods are as follows: When the upper body bend is greater than θ stand , and the bending angular velocity reaches v move , and the encoder position difference is less than θ e , and the product of the left and right angular velocities is greater than v e , switch from standing / walking state to squatting state, When the upper body stops bending during squatting, the angular velocity of the upper body is less than v move , and the product of the left and right angular velocities is less than v e , enter the squat stop state, When squatting or standing up, if the body stands up, the angular velocity of the upper body is less than v move , or the product of the angular velocities is greater than v e , then enter the standing state, When the upper body stops lifting during the process of standing up, the angular velocity of the upper body is greater than v move , or the product of the angular velocities is less than v e When the When the body continues to rise upward in the rising stop state and meets the above-mentioned rising state conditions, the rising state is maintained; when the body angle is less than θ stand , and the upper body angular velocity is less than v move When the mouse is turned on, switch from the standing state to the two-foot standing state / walking state.

[0011] As an optimization, the pitch angle and angular velocity data of the inertial sensor on the user's back and the angle value data of the absolute encoder are collected in real time to identify and judge the motion state and adopt different assistance strategies in different motion states. The control algorithms and executions in different states are as follows: ① If in standing / walking state, execute the walking assistance algorithm, including the zero-force drag algorithm to calculate the torque to overcome gravity and friction, and the torque compensation algorithm based on the hip joint speed to calculate the compensation torque, and add the two to get the final control torque. ② If in the standing state, the assist torque is calculated according to the PD control law of the standing assist, and the final control torque is calculated in combination with the support force compensation. ③ If in the squat stop or stand up stop state, only the control torque of the support force compensation part is calculated.

[0012] As an optimization, the entire control process also includes an error handling mechanism for sensor data jumps, specifically: The data processing module processes the sensor values, including inverting the left hip joint encoder value to unify the left and right directions, performing differential processing on the encoder values ​​of adjacent sampling points and calculating the hip angular velocity, and returning the IMU to zero. At the same time, if the current value changes too much from the values ​​of the previous two sampling points, the variable error is set to 1, indicating a data jump. The state recognition module determines the motion state based on the sensor data. When the error is 1, the state is switched to the initial 0, and the assist motor stops moving to ensure the safety of the wearer. As an optimization, the operation of the motor control module includes motor initialization and sending current values. Specifically, the power assist motor initialization is completed within 3 seconds after the exoskeleton robot is powered on, including turning on the left and right hip power assist motors and sending instructions to set the motor speed limit. Three seconds after power-on, the motor control module only executes the operation of sending the current value, converts the control torque value calculated by the control strategy module into current and sends it to the driver, so that the power-assisting motor operates according to the control torque under the current loop, realizes the exoskeleton's power-assisting action on the human hip, assists the human body to complete bending, getting up, and walking, reduces the burden on the human body, and reduces the risk of muscle fatigue or injury. At the same time, it continuously monitors sensor data and adjusts the control strategy and torque output in real time according to state changes.

[0013] As an optimization, two control modes are included: Follow-up control mode: According to the torque change sensed by the left torque sensor or / and the right torque sensor caused by the instantaneous action of the user, the movement trend of the left or / and right hip of the user is judged, and the left or / and right hip joint assist mechanism is controlled to move along with the movement; 2. Left and right synchronous control mode: The user's torso posture is judged based on the real-time measurement of the angle, angular velocity, angular acceleration and power torque sensed by the back inertial sensor. Then, the angle and angle change of the left and right hip joint power-assisting mechanisms measured by the absolute encoders of the left and right hip joint power-assisting mechanisms are combined to infer the user's current action type, and the left and right hip joint power-assisting mechanisms are controlled to make corresponding actions to complete the power assistance. The detailed judgment is as follows: If the inertial sensor on the back senses that the user's upper body is basically in a vertical state, the left and right hip joint assist mechanisms are alternately bent / extended, and it is judged that the user is in a walking state. The angle amplitude and speed of the hip joint assist mechanisms on the left and right sides of the user are the same, but in opposite directions; If the inertial sensor on the back senses that the user's upper body is leaning forward, the left and right hip joint assist mechanisms will stretch at the same time, and it is judged that the user is standing up. The left and right hip joint assist mechanisms of the user have the same movement angle, speed and direction; If the inertial sensor on the back senses that the user's upper body is leaning to the left, the left hip joint assist mechanism bends and the right hip joint assist mechanism extends, and it is judged that the user is turning left; the left and right hip joint assist mechanisms of the user move in opposite directions, and the angle amplitude and speed of the left hip joint assist mechanism of the user are smaller than the angle amplitude and speed of the right hip joint assist mechanism of the user; If the inertial sensor on the back senses that the user's upper body is leaning to the right, the right hip joint assist mechanism bends and the left hip joint assist mechanism extends, and it is judged that the user is turning right; the left and right hip joint assist mechanisms of the user move in opposite directions, and the angle amplitude and speed of the right hip joint assist mechanism of the user are smaller than the angle amplitude and speed of the left hip joint assist mechanism of the user; If the back IMU sensor senses that the user's upper body is leaning forward, the hip angle value increases and the upper body bending angular velocity reaches the set threshold v move , and the angle difference measured by the absolute encoders at the two hip motors is less than the threshold value θ e And the product of the left and right angular velocities is greater than the threshold v e , determine that the user starts squatting; When the user is in the squatting stop state or the standing stop state, if the back IMU sensor senses that the user's upper body begins to lean back (the angular velocity of the upper body is less than v move or the product of the angular velocities is greater than v e ), to determine that the user has entered the state of standing up; during the process of standing up, the upper body continues to lean back, and the hip angle gradually decreases. The system uses PD control based on position error to calculate the power torque, and combines support force compensation to provide upward power to the user. The power torque is calculated based on the real-time hip angle and angular velocity data, providing greater power to overcome gravity in the early stage of standing up, and the power torque gradually decreases as the body gradually stands upright; when the upper body stops lifting (the upper body angular velocity is greater than v move or the product of the angular velocities is less than v e ) Enters the standing stop state, the system continues to provide support to ensure the stability of the body in the upright position, while continuously monitoring sensor data to determine whether further movement adjustments are needed.

[0014] When the conditions specified in the left-right synchronous control mode are met, control is performed in the left-right synchronous control mode. When the conditions specified in the left-right synchronous control mode are not met, control is performed in the follow-up control mode.

[0015] 1. Hardware system composition The exoskeleton robot includes a back structure, a waist structure, an upper limb wearable structure, a hip joint motion module and a leg binding mechanism. The back structure houses electrical components such as batteries and controllers to provide support and assistance; the waist structure is adjustable to accommodate different body shapes; the upper limb wearable structure uses an inner layer of flexible material combined with an outer layer of rigid load-bearing structure to improve wearing comfort; the hip joint motion module is composed of an assist motor, an encoder, a joint connection plate, a thigh rod, etc. to achieve hip joint motion control; the leg binding mechanism allows the exoskeleton leg rod to fit closely with the human leg.

[0016] The control system includes a motion control board (CSPACE-V4), a torque sensor, an external absolute encoder, an IMU inertial sensor, a lithium battery and a DC-DC power converter.

[0017] 2. Multimodal information collection and processing ①IMU inertial sensor: installed at the center axis of the back frame device, used to detect the spatial posture of the exoskeleton back frame and determine the current exoskeleton gait mode. The sensor uses MEMS technology and integrates a three-axis gyroscope, a three-axis accelerometer and a motion sensor. It achieves accurate measurement and analysis through the independently developed attitude dynamics core algorithm engine and high-dynamic Kalman filter fusion algorithm. It can directly output the inclination angle, with high measurement accuracy and good stability. It can provide real-time data with an update rate of up to 200Hz, meeting various high-precision application requirements. It is widely used in industrial inclination measurement and other fields, has strong anti-electromagnetic interference ability, and can adapt to harsh industrial environments.

[0018] ② Torque sensor: mainly detects the joint torque of the exoskeleton hip and knee joints, rated voltage DC-7.5~15V, rated current ≤150mA, real-time transmission of torque data through RS485 communication, interface protocol is RS-485 bus, baud rate 115200, elastomer material is stainless steel, measurement range ±300N.m, comprehensive measurement accuracy ±0.5%FS, creep 0.1%F.S30min, working environment temperature 0℃~+45℃, storage and transportation environment temperature -20℃~+85℃.

[0019] ③Encoder: In addition to the internal incremental encoder, each power-assisted motor has an external absolute encoder with a rated voltage of DC-7.5~15V, a rated current ≤150mA, RS485 communication, an interface protocol RS-485 bus, a default baud rate of 115200, a measurement angle range of 0~360°, a resolution of 14Bit (0.0219° / LSB), a maximum speed of 1200rmp, a power-on time of 100ms, an input shaft allowable deviation of <0.3mm, a magnetic flux of ±30mT...±70mT, and a typical distance between the magnet and the encoder chip of 0.5mm-2.5mm.

[0020] 3. Control algorithm implementation State machine design and switching logic: The state machine starts from the standing / walking state, including the squatting state, squatting stop state, standing up state, and standing up stop state. The switching conditions are based on multi-modal information such as the body pitch angle measured by the IMU on the back of the body, the body pitch angular velocity, the angle value measured by the absolute encoder at the two hip motors and the product of its angular velocity. For example, the condition for the two-foot standing state / walking state → squatting state is that the upper body bending is greater than θ stand , and the bending angular velocity reaches v move , and the encoder position difference is less than θ e , and the product of the left and right angular velocities is greater than v e .

[0021] ①Zero-force dragging algorithm based on torque compensation (applicable to standing / walking state) The robot driver works in current / torque mode to control the torque , where the gravity term Simplified to the joint angle i Trigonometric functions sin(θ) Related functions t G =Ksin(θ) , friction torque at the joint , thus obtaining the control torque of the knee and hip joints during swing t knee and t hip , controller parameters are determined by mass, center of mass estimation and actual machine test tuning K i (i=1,2…7) to counteract gravity and achieve smooth human-machine interaction force.

[0022] ②Torque compensation algorithm based on hip joint velocity (applicable to standing / walking state) Two sampling intervals are set. The large sampling interval (greater than 10 sampling points) is used to determine the acceleration or deceleration process of the hip joint movement and determine the direction of the control torque. The small sampling interval (less than 5 sampling points) is used to calculate the torque compensation value, compensate the hip torque according to the current angle and speed, and enhance the exoskeleton's tracking performance.

[0023] ③PD control based on position error (applicable to standing state) Using PD control based on position error, the control law is:

[0024] in e(t)=θ d - i is the error, where i d is the desired angle, i It is the motor angle value measured in real time. K P is the proportional phase coefficient, K D is the differential phase coefficient. When assisting in getting up, the desired angle can be set to 0. K P and K D value to obtain the appropriate assist torque.

[0025] ④Support force compensation (applicable to squat stop state, stand up stop state and stand up state) The part of the exoskeleton robot above the hip is considered as the distance from the hipl Meter, mass m When the human body is stationary, the torque of the upper part of the hip support robot is T=mlsin(θ) , the motor provides the corresponding ratio K Torque u=KT , can be adjusted according to actual needs in different states K value.

[0026] 4. Software system functions The design of key control components includes motor control (involving CAN address and baud rate settings, CAN communication protocol, etc.), main control board, IMU's RS485 communication protocol (address and baud rate settings, inclination RS485 communication protocol, etc.), torque sensor RS485 communication protocol (write address, clear, read torque value and other instructions and corresponding data format), encoder RS485 communication protocol (set baud rate, communication address, clear angle value, read angle value and other instructions and corresponding data format).

[0027] The control algorithm design covers algorithm logic (state machine status and control logic), control algorithm (squat assistance algorithm and walking assistance algorithm) and algorithm programming (consisting of serial port initialization, sensor transmission and reception, control and other modules; the control part includes data processing, state recognition, control strategy and motor control module).

[0028] Beneficial effects of the present invention ① Through the fusion of multiple sensors such as IMU, encoder and torque sensor, the wearer's body posture, joint force and movement status can be accurately and comprehensively obtained, so as to accurately identify the human body's movement status and provide appropriate assistance at the right time.

[0029] ② Through multi-sensor fusion and appropriate control algorithms, the exoskeleton robot can judge the real-time movement state of the human body and adjust the strength and direction of assistance, ensure the flexibility of human-machine collaboration, and reduce discomfort during human-machine interaction.

[0030] ③ By running corresponding assistance algorithms for different action states such as squatting, walking, etc., the control accuracy of the exoskeleton is significantly improved, enabling the exoskeleton robot to cope with complex and changing working environments and effectively reducing the risk of muscle fatigue and injury for the wearer. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 The overall mechanical structure of the exoskeleton Figure 2 Flow chart of the control method Figure 3 State determination flow chart DETAILED DESCRIPTION

[0032] The present invention relates to a hip-assisted exoskeleton control method based on force-position information fusion, which aims to fuse multiple types of sensor data, accurately determine the wearer's motion state, and then provide adaptive assistance, enhance the wearer's sports performance and improve work efficiency. The following is a detailed explanation of the specific implementation of the present invention: 1. Hardware system composition like Figure 1 As shown, the exoskeleton robot includes a back structure 1, a waist structure 2, an upper limb wearable structure 3, a hip joint motion module 4 and a leg binding mechanism 5. The back structure 1 accommodates electrical components such as batteries and controllers to provide support and assistance; the waist structure is adjustable to adapt to different body shapes; the upper limb wearable structure 3 combines an inner layer of flexible material with an outer layer of rigid bearing structure to improve wearing comfort; the hip joint motion module 4 is composed of an assist motor, an encoder, a joint connection plate, a thigh rod, etc. to achieve hip joint motion control; the leg binding mechanism 5 makes the exoskeleton leg rod fit closely with the human leg.

[0033] The control system includes a motion control board (CSPACE-V4), a torque sensor, an external absolute encoder, an IMU inertial sensor 6, a lithium battery and a DC-DC power converter.

[0034] S1. Multimodal information collection and preprocessing IMU data processing: The IMU on the back of the body is used to measure the body pitch angle θ body and the body pitch angular velocity v body , the positive direction is the direction in which the body bends downward. Perform zeroing processing so that the IMU angle value is 0 when standing upright, and the bending direction is the positive direction. Send the left and right hip angles, angular velocity values, and the IMU angle and angular velocity values ​​in the pitch direction to the host computer for monitoring. At the same time, add a data error reporting function. If the current value changes too much from the values ​​of the previous two sampling points (such as by judging whether the angle exceeds a reasonable range in the program), set the variable error to 1, and send the variable value to the host computer to monitor the danger. The host computer can handle and cancel the error.

[0035] Encoder data processing: The absolute encoders at the two hip motors are used to measure the angle value θ of the right hip right The angle θ with the left hip left , the positive direction is the direction of the thigh kicking forward. The left hip joint encoder value is inverted so that the direction of the left and right hips lifting forward is the positive direction. Then the encoder values ​​of two adjacent sampling points are differentially processed, and the right hip angular velocity v is calculated based on the sampling time. right , left hip angular velocity v left .

[0036] Torque sensor data processing: Torque sensors are installed at the output ends of the hip drive motors on both sides of the hip joint exoskeleton to obtain the human-computer interaction force between the wearer and the exoskeleton in real time, providing a basis for accurately controlling the size of the assistance.

[0037] S2. The specific implementation method of motion state recognition and judgment is as follows: Threshold setting: Standing body angle threshold θ stand =15; moving body angular velocity threshold v move =15; standing hip angle difference threshold θ hip =15.

[0038] ① Initial state state=0: After the system is powered on, it first enters the initial state, at which time the output value of the state recognition module is 0. In this state, the system performs initialization operations, including sensor calibration, motor initialization, etc. At the same time, it waits for the sensor data to stabilize in preparation for subsequent accurate identification of the motion state.

[0039] ② Standing / walking state state=1: When the following conditions are met, the system determines that it has entered the standing / walking state, and the output value of the state recognition module is 1.

[0040] Stand up straight, that is, θ body <θ stand ; The back stops moving and the hips stop moving, that is, v body <v move And θ body <θ stand .

[0041] ③Squatting state state=2: When θ body >10, v body >v move 、v hip_product >50,v left >15,v right When the value is >15, the system determines that it has entered the squatting state, and the output value of the state recognition module is 2.

[0042] In this state, the squatting action mainly relies on gravity, but the system will continue to monitor sensor data to prepare for possible subsequent assistive actions. At the same time, based on the mechanical characteristics of the squatting process, it will preliminarily estimate the timing and size of the assist that may be needed, such as preparing to provide a certain amount of buffering assistance when the squatting limit is about to be reached.

[0043] ④ Squat stop state=3: When the upper body stops bending during the squat process, that is, v body <15 and v hip_product When ≤400, the system determines that it has entered the squat stop state, and the output value of the state recognition module is 3.

[0044] In this state, although the body temporarily stops squatting, the system still keeps monitoring the sensor data because the exoskeleton may be needed to provide a certain amount of support to maintain the stability of the body in the squatting stop position to prevent instability caused by excessive squatting or sudden standing up.

[0045] ⑤Standing up state=4: When the body stands up, v body <-13 or v hip_product >450,v right <-5 or v left <-5, the system determines that it has entered the standing up state, and the state recognition module outputs a value of 4.

[0046] The process of getting up requires assistance. In this state, the system starts the assistance algorithm to provide the wearer with upward assistance according to the preset control strategy. At the same time, the assistance is adjusted in real time according to the speed and angle of the body getting up to ensure that the assistance effect matches the wearer's exercise needs. For example, a large amount of assistance is provided at the beginning of getting up to overcome gravity, and the assistance is gradually reduced as the body gradually stands upright.

[0047] ⑥When the upper body stops lifting, that is, v body >v move or v hip_product <v e , the system determines that it has entered the stand-up stop state, and the output value of the state recognition module is 5.

[0048] At this time, although the standing up action stops temporarily, the system still needs to maintain a certain amount of assistance or support to prevent the body from suddenly sinking or shaking due to the loss of assistance. At the same time, the system determines whether further action adjustments are needed based on the current posture and sensor data, such as entering the standing up state again or switching to other states.

[0049] ⑦Error state state=0: When the left and right hip angles are out of range (less than 40 or greater than 150), or the body angle is out of range (less than 40 or greater than 150), or the error is 1, switch to the initial state.

[0050] S3. Assist in strategy implementation: ① Assistance in standing / walking state: In the standing / walking state (move=1), the assistance strategy mainly includes reducing the influence of friction and the gravity of the leg rods on the wearer's movement and compensating the hip torque according to the current angle and speed.

[0051] The zero-force drag algorithm uses a zero-force control strategy based on torque compensation to enable the robot drive to operate in current / torque mode. This strategy precisely controls the servo motor to provide gravity and friction compensation required for robot movement. ,in is the gravity term in the robot dynamics equation, which can be simplified to the joint angle in this exoskeleton robot i Trigonometric functions sin(θ) The function of t G =Ksin(θ) ; Friction torque at the joint (Consists of Coulomb friction and viscous friction). From this, we can get the control torque of the knee and hip joints during swinging: t knee and t hip , by adjusting the controller parameters K i (i=1,2…7) (mass and center of mass estimation and actual machine testing can be used to adjust parameters) to achieve smoother human-computer interaction and offset gravity.

[0052] The torque compensation algorithm based on hip joint velocity sets two sampling intervals. The large sampling interval (usually greater than 10 sampling points) is used to determine the acceleration or deceleration process of the hip joint movement and determine the direction of the control torque; the small sampling interval (usually less than 5 sampling points) is used to calculate the value of the torque compensation to enhance the robot's follow-up performance. The specific implementation can determine the calculation method of the compensation torque based on the monitoring and calculation method of the speed and angle changes in the program, for example, based on v right 、v left Speed ​​data determines acceleration or deceleration, according to θ right ,θ left Calculate compensation value based on equal angle data.

[0053] ② Squat state assist (preparation phase): In the squat state (move=2), although squatting mainly relies on gravity, the system will body 、v right 、v left etc.), angle (θ body ,θ right ,θ left etc.) and the preset mechanical model to estimate the force conditions during squatting.

[0054] When approaching the squat limit position (which can be judged by angle data, such as θ bodyWhen the value approaches a certain value, the system is ready to start buffering assistance to prevent discomfort or injury caused by excessive gravity. At this time, the system begins to adjust the output torque of the motor so that it can provide a certain reverse resistance at the right time, slow down the squatting speed, and achieve a smooth squatting.

[0055] ③ Squat stop state assistance (support phase): In the squat stop state (move=3), the system provides support force to maintain the stability of the body in the squat stop position.

[0056] The part of the exoskeleton robot above the hip is considered as a distance from the hip. l The mass of rice is m When the human body is stationary, the torque of the upper part of the hip support robot is T=mlsin(θ) (in i is the back pitch angle), the motor should provide a corresponding proportion of support force. If the proportional coefficient is K The torque of u=KT , in this way, it provides stable support for the wearer and reduces the burden on the leg muscles.

[0057] ④ Standing up state assistance (main assistance stage): Standing up assistance adopts PD control based on position error. The control law is:

[0058] in e(t)=θ d - i Error i d is the desired angle, i is the motor angle value measured in real time, K P is the proportional phase coefficient, K D is the differential phase coefficient. When performing standing assistance, the desired angle can be set to 0.

[0059] The system is based on the real-time hip angle (θ right ,θ left ) and angular velocity (v right 、v left ) data, calculate the current position error, and adjust the proportional phase coefficient K P and differential phase coefficient K D , providing appropriate assist torque. In the early stage of standing up, the position error is large, and the system provides a large assist torque to overcome gravity and start the standing up action; as the body gradually approaches the upright state, the position error decreases, and the assist torque also decreases accordingly, until the upright position is reached, the assist torque gradually drops to zero, achieving a smooth transition.

[0060] ⑤ Assistance in the standing-up and stopping state (stable holding stage): In the standing-up and stopping state (move=5), the system continues to provide a certain amount of support force to ensure the stability of the body in the upright position.

[0061] The support force calculation method similar to the squat stop state is also used, according to the current body posture (θ body ,θ right ,θ left The system calculates the required support torque based on the relationship between the sensor data and the mechanical state, and outputs the corresponding force through the motor to prevent the body from shaking or sinking due to the loss of support. At the same time, the system continuously monitors the sensor data to determine whether further action adjustments are needed, such as re-entering the walking state (move=1) or other actions. In the program, the calculation and output of the support force are adjusted in real time according to the changes in the state and data.

[0062] S3. Motor control: ① Motor initialization: The motor control module completes the motor initialization operation within 3 seconds after the system is powered on. Turn on the left and right hip motors, and send corresponding instructions to set the motor speed limit. These initialization settings ensure that the motor can work stably and safely during subsequent operation, and can accurately output the required torque and speed according to the control strategy.

[0063] ②Send current value: 3 seconds after power-on, the motor control module mainly performs the operation of sending current value. After converting the control torque value calculated by the control strategy module into current, it is sent to the driver. The driver controls the left and right hip assist motors to operate under the current loop according to the received current value, so that the torque output by the motor matches the calculated control torque value, thereby achieving precise control of the exoskeleton and providing appropriate assistance to the wearer. During the whole process, the motor control module continuously adjusts the control torque according to the sensor data (such as θ right ,θ left 、v right 、v left ,θ body 、v body The exoskeleton’s power-assisting effect is always consistent with the wearer’s movement needs by adjusting the control strategy and updating the sent current value in real time.

Claims

1. A hip-assisted exoskeleton control method based on force-position information fusion, comprising the following steps: S1. Multimodal information collection and preprocessing The inertial sensor installed on the back is used to collect the body pitch angle and pitch angular velocity information, the absolute encoder is used to collect the hip joint angle information, and the torque sensor is used to collect the hip joint torque information. The collected inertial sensor data is filtered to remove noise interference, and the absolute encoder data is calibrated to ensure the accuracy of angle measurement. The torque sensor data is corrected for zero drift. S2. Motion state recognition and judgment The body pitch angle and angular velocity collected by the inertial sensor are combined with the angle value of the absolute encoder to make state machine switching judgments according to preset threshold conditions. The states include standing / walking state, squatting state, squatting stop state, standing up state, and standing up stop state. Set the body pitch angle threshold θ stand , body pitch angular velocity threshold v move , hip position difference threshold θ e , hip angular velocity product threshold v e The motion states include standing / walking state, squatting state, squatting stop state, standing up state, and standing up stop state. The specific judgment methods are as follows: When the upper body bend is greater than θ stand , and the bending angular velocity reaches v move , and the encoder position difference is less than θ e , and the product of the left and right angular velocities is greater than v e , switch from standing / walking state to squatting state, When the upper body stops bending during squatting, the angular velocity of the upper body is less than v move , and the product of the left and right angular velocities is less than v e , enter the squat stop state, When squatting or standing up, if the body stands up, the angular velocity of the upper body is less than v move , or the product of the angular velocities is greater than v e , then enter the standing state, When the upper body stops lifting during the process of standing up, the angular velocity of the upper body is greater than v move , or the product of the angular velocities is less than v e When the When the body continues to rise upward in the rising stop state and meets the above-mentioned rising state conditions, the rising state is maintained; when the body angle is less than θ stand , and the upper body angular velocity is less than v move , switch from the standing state to the two-foot standing state / walking state.

2. The hip-assisted exoskeleton control method based on force-position information fusion according to claim 1 is characterized in that: The pitch angle and angular velocity data of the inertial sensor on the user's back and the angle value data of the absolute encoder are collected in real time to identify and judge the motion state and adopt different assistance strategies in different motion states. The control algorithms and executions in different states are as follows: ① If in standing / walking state, execute the walking assistance algorithm, including the zero-force drag algorithm to calculate the torque to overcome gravity and friction, and the torque compensation algorithm based on the hip joint speed to calculate the compensation torque, and add the two to get the final control torque. ② If in the standing state, the assist torque is calculated according to the PD control law of the standing assist, and the final control torque is calculated in combination with the support force compensation. ③ If in the squat stop or stand up stop state, only the control torque of the support force compensation part is calculated.

3. The hip-assisted exoskeleton control method based on force-position information fusion according to claim 1, characterized in that: The entire control process also includes an error handling mechanism for sensor data jumps, specifically: The data processing module processes the sensor values, including inverting the left hip joint encoder value to unify the left and right directions, performing differential processing on the encoder values ​​of adjacent sampling points and calculating the hip angular velocity, and returning the IMU to zero. At the same time, if the current value changes too much from the values ​​of the previous two sampling points, the variable error is set to 1, indicating a data jump. The state recognition module determines the motion state based on the sensor data. When the error is 1, the state is switched to the initial 0, and the power assist motor stops moving to ensure the safety of the wearer.

4. The hip-assisted exoskeleton control method based on force-position information fusion according to claim 1, characterized in that: The operation of the motor control module includes motor initialization and sending current values, specifically: The power-assist motor initialization is completed within 3 seconds after the exoskeleton robot is powered on, including enabling the left and right hip power-assist motors and sending instructions to set the motor speed limit. Three seconds after power-on, the motor control module only executes the operation of sending the current value, converts the control torque value calculated by the control strategy module into current and sends it to the driver, so that the power-assisting motor operates according to the control torque under the current loop, realizes the exoskeleton's power-assisting action on the human hip, assists the human body to complete bending, getting up, and walking, reduces the burden on the human body, and reduces the risk of muscle fatigue or injury. At the same time, it continuously monitors sensor data and adjusts the control strategy and torque output in real time according to state changes.

5. The hip-assisted exoskeleton control method based on force-position information fusion according to claim 1 includes two control modes:

1. Follow-up control mode: According to the torque change sensed by the left torque sensor or / and the right torque sensor caused by the instantaneous action of the user, the movement trend of the left or / and right hip of the user is judged, and the left or / and right hip joint assist mechanism is controlled to move along with the movement; 2. Left and right synchronous control mode: The user's torso posture is judged based on the real-time measurement of the angle, angular velocity, angular acceleration and power torque sensed by the back inertial sensor. Then, the angle and angle change of the left and right hip joint power-assisting mechanisms measured by the absolute encoders of the left and right hip joint power-assisting mechanisms are combined to infer the user's current action type, and the left and right hip joint power-assisting mechanisms are controlled to make corresponding actions to complete the power assistance. The detailed judgment is as follows: If the inertial sensor on the back senses that the user's upper body is basically in a vertical state, the left and right hip joint assist mechanisms are alternately bent / extended, and it is judged that the user is in a walking state. The angle amplitude and speed of the hip joint assist mechanisms on the left and right sides of the user are the same, but in opposite directions; If the inertial sensor on the back senses that the user's upper body is leaning forward, the left and right hip joint assist mechanisms will stretch at the same time, and it is judged that the user is standing up. The left and right hip joint assist mechanisms of the user have the same movement angle, speed and direction; If the inertial sensor on the back senses that the user's upper body is leaning to the left, the left hip joint assist mechanism bends and the right hip joint assist mechanism extends, and it is judged that the user is turning left; the left and right hip joint assist mechanisms of the user move in opposite directions, and the angle amplitude and speed of the left hip joint assist mechanism of the user are smaller than the angle amplitude and speed of the right hip joint assist mechanism of the user; If the inertial sensor on the back senses that the user's upper body is leaning to the right, the right hip joint assist mechanism bends and the left hip joint assist mechanism extends, and it is judged that the user is turning right; the left and right hip joint assist mechanisms of the user move in opposite directions, and the angle amplitude and speed of the right hip joint assist mechanism of the user are smaller than the angle amplitude and speed of the left hip joint assist mechanism of the user; If the back IMU sensor senses that the user's upper body is leaning forward, the hip angle value increases and the upper body bending angular velocity reaches the set threshold v move , and the angle difference measured by the absolute encoders at the two hip motors is less than the threshold value θ e And the product of the left and right angular velocities is greater than the threshold v e , determine that the user starts squatting; When in the squatting stop state or the standing up stop state, if the IMU sensor on the back senses that the user's upper body begins to lean backward, it is judged that the user has entered the standing up state; during the standing up process, the upper body continues to lean backward and the hip angle gradually decreases. The system uses PD control based on position error to calculate the assist torque, and combines support force compensation to provide upward assistance to the user. The assist torque is calculated based on the real-time hip angle and angular velocity data, providing greater assistance to overcome gravity in the early stage of standing up, and the assist torque gradually decreases as the body gradually stands upright; when the upper body stops lifting and enters the standing up stop state, the system continues to provide support force to ensure the stability of the body in the upright position, and continuously monitors the sensor data to determine whether further movement adjustments are needed.

Images