Exoskeleton robot adaptive cooperative control method

Abstract

This invention discloses an adaptive collaborative control method for exoskeleton robots, belonging to the field of exoskeleton robot control technology. The method includes: collecting the target joint motion state, human-machine interaction force, and exoskeleton joint output torque within the current handling cycle to obtain predicted joint motion state and human-machine interaction force change rates; performing directional unification and normalization processing on the predicted joint motion trend and human-machine interaction force change trend to calculate a protective force enhancement characteristic quantity; calculating the reinjection amplification risk increment based on the protective force enhancement characteristic quantity, human-machine interaction force change rate, and exoskeleton joint output torque change rate; constraining the normal expected assist torque based on the reinjection amplification risk increment to obtain the control torque and drive the exoskeleton joint to output assist. This invention can avoid misjudging the human protective force enhancement induced by load disturbance as an increase in active assist demand, reducing the risks of assist overshoot, human-machine aggression, and handling posture instability.

Description

Technical Field

[0001] This invention relates to the field of exoskeleton robot control technology, and in particular to an adaptive cooperative control method for exoskeleton robots. Background Technology

[0002] Exoskeleton robots are intelligent assistive devices that coordinate with human limbs or torso through mechanical support structures, joint drive mechanisms, and sensor control systems. They are widely used in high-intensity work scenarios such as industrial handling, logistics loading and unloading, warehousing sorting, workpiece transfer, and lifting and lowering heavy objects. In industrial handling, the wearer typically needs to perform continuous movements such as bending, lifting, turning, moving forward, and lowering. The exoskeleton robot needs to output assist torque in a timely manner based on the human's movement and human-robot interaction status to reduce the load on target joints such as the waist, hips, knees, shoulders, or elbows. Existing exoskeleton robots typically use inertial measurement units, joint encoders, force sensors, or joint torque sensors to collect signals such as joint posture, joint velocity, human-robot interaction forces, and output torque. They then employ position control, force control, or impedance control methods to enable the exoskeleton joints to output corresponding assistance following the human's movement trends.

[0003] However, in real-world industrial material handling scenarios, the objects being handled are not always in a stable stress state. Goods may experience brief periods of jamming, sudden release, partial unloading, or shifting of the center of gravity due to pallet edges, shelf gaps, packaging deformation, floor obstacles, or material movement within boxes. In such cases, the wearer's true intention may not be to further enhance the handling action, but rather to prevent the goods from slipping, causing body imbalance, or joint impact, resulting in protective force-increasing responses such as gripping, torso tension, joint resistance, or short-term braking. These protective responses cause a rapid increase in human-machine interaction forces at the human-machine interface, which can easily be confused with the wearer's active demand for increased handling assistance in sensor readings. If the exoskeleton robot judges the assistance demand solely based on increased human-machine interaction forces or single joint movement signals, it will be difficult to distinguish between the enhanced active handling action of the human and the protective force-increasing response induced by load disturbances.

[0004] Existing adaptive collaborative control methods for exoskeleton robots still have shortcomings under complex load disturbance conditions: when load disturbances trigger human protective force amplification, the increased human-robot interaction force may be misinterpreted by the control system as an increased demand for active assistance, causing the exoskeleton robot to continue increasing its assistance torque or impedance stiffness. This results in the external load disturbance, human protective force amplification, and exoskeleton assistance gain amplifying each other along the same force chain, leading to assistance overshoot, human-robot aggression, or instability in the handling posture. Especially in industrial material handling, load jamming release, partial unloading, and center of gravity shift are sudden and short-lived. Relying solely on preset trajectories, single force feedback, or ordinary motion intention recognition is insufficient to promptly determine the true source of the increased interaction force, affecting the compliance, safety, and collaborative stability of the exoskeleton robot during handling operations. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies in the collaborative control of exoskeletons for industrial handling, which make it difficult to distinguish between the protective force increase induced by load disturbance and the increased demand for active assistance from the human body. This leads to the exoskeleton robot easily misinterpreting the increase in human-machine interaction force as an increase in active assistance demand, resulting in assistance overshoot, human-machine conflict and instability in handling posture. Therefore, this invention proposes an adaptive collaborative control method for exoskeleton robots.

[0006] To address the problems existing in the prior art, the present invention adopts the following technical solution: Adaptive cooperative control methods for exoskeleton robots include: S1. Collect human-machine collaboration status data within the current handling cycle to obtain the target joint motion status, human-machine interaction force, and exoskeleton joint output torque; S2. Obtain the predicted joint motion state based on the target joint motion state, and obtain the human-computer interaction force change rate based on the human-computer interaction force; S3. Obtain the protective force enhancement characterization quantity based on the predicted joint motion state and the human-machine interaction force change rate; S4. Obtain the reinjection amplification risk increment based on the protective force enhancement characterization quantity, the human-machine interaction force change rate, and the exoskeleton joint output torque; S5. The control torque is obtained based on the predicted joint motion state, the target joint motion state, and the reperfusion amplification risk increment. S6. Drive the exoskeleton joint to output assistance according to the control torque, and perform safety closed-loop correction based on the human-machine collaboration status data at the next control moment.

[0007] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention collects the target joint motion state, human-machine interaction force, and exoskeleton joint output torque. Combined with the predicted rate of change of human-machine interaction force after unifying the joint motion state and direction, a protective force enhancement representation quantity is constructed. This allows the controller to determine whether the increase in human-machine interaction force is opposite to the wearer's predicted joint motion trend. When goods become stuck, partially unloaded, the center of gravity shifts, or the load sways during industrial handling, the wearer's protective grip, torso tension, joint resistance, or short-term braking response can be identified through the protective force enhancement representation quantity. This avoids directly identifying such protective force enhancement as an increase in active assistance demand. It improves the exoskeleton robot's ability to recognize the true collaborative state of the human body under complex handling conditions, making the assistance output no longer solely dependent on changes in the magnitude of human-machine interaction force. This helps reduce assistance overshoot and human-machine aggression caused by misjudging assistance demand.

[0008] 2. This invention further calculates the incremental risk of backflow amplification based on the protective force enhancement characteristic quantity, the rate of change of human-machine interaction force, and the rate of change of exoskeleton joint output torque. When the risk of backflow amplification increases, the normal expected assist torque is constrained, so that the exoskeleton robot no longer continues to amplify the assist output along the direction of increasing human-machine interaction force, thereby blocking the process of continuous amplification of load disturbance, human protective force enhancement, and exoskeleton assist gain along the same force chain. Through gradual recovery and safety closed-loop correction, normal impedance assist is smoothly restored when the risk of backflow amplification decreases, and the assist output is reduced or safety shutdown protection is triggered when the risk continues to increase or the human-machine interaction state is unstable. This can improve the compliance, safety, and collaborative stability of industrial handling exoskeleton robots in sudden load disturbance scenarios and reduce the risk of handling posture instability. Attached Figure Description

[0009] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating an adaptive cooperative control method for an exoskeleton robot according to an embodiment of the present invention. Detailed Implementation

[0010] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0011] Example 1, see Figure 1 This embodiment provides an adaptive cooperative control method for exoskeleton robots, specifically including: This embodiment uses a waist-assisted exoskeleton robot to assist a wearer in carrying box-shaped goods as an example. The exoskeleton robot includes a waist support component, a leg support component, a hip joint assist component, a waist flexion and extension assist component, a joint actuator, an inertial measurement unit, a joint encoder, a force sensor, a joint torque sensor, and a controller. The inertial measurement unit is located at the waist and back support parts and the thigh support parts of the wearer's torso to collect posture and acceleration data. The joint encoder is located at the hip joint assist component and the waist flexion and extension assist component to collect target joint angles. The force sensor is located at the waist strap, leg strap, or human-machine interface to collect human-machine interaction forces. The joint torque sensor is located on the transmission path of the assist joint to collect the exoskeleton joint output torque. The controller is used to execute the adaptive cooperative control method of this embodiment. In this embodiment, the target joint is an exoskeleton assistive joint that participates in lifting, lowering, or transferring actions during industrial handling. For a lumbar assistive exoskeleton, the target joint includes the lumbar flexion-extension joint and the hip joint. For an upper limb assistive exoskeleton, the target joint may include the shoulder joint and the elbow joint. For ease of explanation, this embodiment uses a single target joint as an example to describe the calculation process. When the exoskeleton robot includes multiple target joints, the same data acquisition, orientation unification, normalization, trend prediction, risk assessment, and control torque generation processes are performed on each target joint, or the state variables of multiple target joints are combined into a vector and then the same type of calculation is performed.

[0012] The method in this embodiment includes the following steps; S1. Human-machine collaboration status data collection and synchronization; Specifically, the controller collects human-machine collaboration status data between the exoskeleton robot and the wearer during the current handling cycle; the current handling cycle refers to the time interval from when the wearer begins to perform a lifting, transporting, or lowering action to when the action ends; for continuous handling operations, the starting point of the handling cycle can be determined based on the moment when the target joint angular velocity changes from stationary to non-stationary, and the ending point of the handling cycle can be determined based on the moment when the target joint angular velocity returns to stationary and the human-machine interaction force tends to stabilize. Human-machine collaboration status data includes target joint angles Target joint angular velocity Target joint angular acceleration Human-computer interaction and exoskeleton joint output torque ; in, The target joint angle at the current control time t is used to characterize the joint posture of the wearer and the exoskeleton robot at this assisted joint. This represents the angular velocity of the target joint at the current control time t, used to characterize the speed and direction of the target joint's movement. This represents the target joint angular acceleration at the current control time t, used to characterize the changing trend of the target joint's motion state; The force representing the human-machine interaction at the current control time t is used to characterize the interaction force between the wearer and the exoskeleton robot at the human-machine interface. This represents the output torque of the exoskeleton joint at the current control time t, used to characterize the actual assist torque output by the exoskeleton robot to the target joint; In one implementation, the target joint angle q(t) is directly acquired by a joint encoder; the target joint angular velocity... The target joint angles are obtained by processing the continuously output target joint angles over time from the joint encoder, or by transforming the angular velocity data output from the inertial measurement unit; the target joint angular acceleration is obtained. The force is obtained by processing the target joint angular velocity over time, or by converting the acceleration data output by the inertial measurement unit into the geometric relationship of the exoskeleton linkage; human-machine interaction force. Force sensors located in the waist straps, leg straps, human-machine interface pads, or joint transmission parts are used to collect data; exoskeleton joint output torque. It is directly acquired by the joint torque sensor; when the joint torque sensor is not installed, it can also be calculated based on the driver current, motor torque constant, gear ratio and transmission efficiency. To ensure that subsequent calculations use the same human-machine collaboration state, the controller adjusts the target joint angle... Target joint angular velocity Target joint angular acceleration Human-computer interaction and exoskeleton joint output torque Sampling time synchronization is performed; specifically, the controller reads data from each sensor at a unified control cycle; if different sensors have different sampling frequencies, the controller's control time is used as a reference, the most recent sampled value or moving average of high-frequency data is taken, and time alignment processing is performed on low-frequency data to obtain human-machine collaboration status data corresponding to the same control time t.

[0013] S2. Predict joint motion state, and unify the processing of human-computer interaction force change rate and direction; Specifically, the controller obtains the predicted joint motion state based on the target joint motion state; the predicted joint motion state includes predicting joint angles. And predicting joint angular velocity ;in, This represents the short-term prediction window. The short-term prediction window is used to predict the movement trend of the wearer's target joint before the exoskeleton robot actually outputs assistance. In this embodiment, the short-term prediction window can be between 100ms and 200ms, which matches the control requirements of the exoskeleton robot to predict human movements in advance. For upper limb exoskeletons with faster response speeds, a shorter short-term prediction window can be used. For waist or lower limb assist exoskeletons, a longer short-term prediction window can be used. In one implementation, the controller determines the target joint angle based on the current control moment. Target joint angular velocity and target joint angular acceleration Short-time motion extrapolation is used to calculate and predict joint angles. And predicting joint angular velocity : in, This indicates the predicted joint angle after the short-term prediction window; This indicates the predicted joint angular velocity after the short prediction window; Indicates the target joint angle at the current control moment; This represents the target joint angular velocity at the current control moment; This represents the target joint angular acceleration at the current control moment; Indicates a short-term forecast window; The above two formulas are derived from the kinematic relationship of short-time uniform acceleration. Since the control cycle of the exoskeleton robot is short, the angular acceleration of the target joint can approximately characterize the trend of joint motion change in a short time between adjacent control moments. Therefore, the above formulas can be used to predict the joint motion state without introducing a complex training model.

[0014] To enable comparison of the predicted joint angular velocity and the rate of change of human-machine interaction force under the same directional reference, the controller performs directional unification processing on the human-machine interaction force and the predicted joint angular velocity before calculating the protective force enhancement quantity. Specifically, the positive direction of assistance for the target joint is predefined as the direction of rotation of the target joint in which the exoskeleton robot assists the wearer in completing the current carrying action. When the target joint is a lumbar flexion-extension joint, the positive direction of assistance is the direction in which the trunk is assisted to return from a flexed state to an upright state. When the target joint is a hip joint, the positive direction of assistance is the direction in which the thigh is assisted to complete the lifting or supporting action relative to the trunk. When the target joint is a shoulder or elbow joint, the positive direction of assistance is the direction in which the upper limb is assisted to support the load. The controller determines the projection component of the human-machine interaction force in the positive direction of the target joint assistance based on the target joint structure and the force sensor installation orientation, and uses this projection component as the orientation-unified human-machine interaction force; the orientation-unified human-machine interaction force is denoted as... If the force sensor is directly installed in the direction of the target joint's assistance, then Equal to human-computer interaction If the force sensor's installation direction forms an angle with the target joint's assist direction, the controller will adjust the human-machine interaction force according to the force sensor's installation orientation and the target joint's linkage direction. Projecting onto the positive direction of the target joint assist, we obtain... ; When the human-computer interaction force is the scalar force value measured by the force sensor, and the angle between the force sensor measurement direction and the positive direction of the target joint assistance is... At that time, the human-computer interaction force after unifying the direction is obtained according to the following relationship: in, This represents the human-computer interaction force after the direction is unified. This indicates the human-computer interaction force collected by the force sensor. This indicates the angle between the force sensor's measurement direction and the positive direction of the target joint's assistance; The rate of change of human-computer interaction force after unifying the direction is calculated as follows: in, The rate of change of human-computer interaction force after unification of direction; This represents the human-computer interaction force after the direction is unified at the current control moment; This represents the human-computer interaction force after the direction is unified at the previous sampling time. This represents the sampling time interval between two adjacent sampling times; The controller will predict the joint angular velocity. The sign of the predicted joint angular velocity is obtained by unifying the sign according to the positive direction of the target joint assist. When predicting joint angular velocity When the positive direction of the target joint assists the target joint, When predicting joint angular velocity When the positive direction is opposite to the positive direction of the target joint assist, ; Through the above-mentioned directional unification process, the predicted joint angular velocity after directional unification is obtained. The rate of change of human-computer interaction force after unification of direction Both directions are taken as positive direction with the positive direction of the target joint assistance, so that the opposite relationship between the two can be used to determine whether the increase in interactive force is opposite to the predicted joint movement trend.

[0015] S3. Normalization processing and calculation of protective force enhancement characteristics; Specifically, the controller predicts the joint angular velocity based on the unified direction. The rate of change of human-computer interaction force after unification of direction Obtain protective force characterization quantity Protective force enhancement characterization quantity This is used to characterize whether the current increase in human-computer interaction force is more likely to come from the protective force enhancement induced by load disturbance, rather than from the increased need for active assistance from the wearer; In industrial material handling, if the wearer actively continues to lift or transfer goods, the joint movement trend indicated by the predicted joint angular velocity usually aligns with the increasing trend of human-machine interaction force. For example, when the wearer actively lifts goods, the joint movement trend is to continue lifting, and the increased interaction force can be interpreted as an enhancement of the active handling action. Conversely, if the goods are jammed and released, the center of gravity shifts, or partial unloading causes external load disturbances, the wearer may instinctively tighten their torso, grip the load, or brake the joints. In this case, the human-machine interaction force increases, but the predicted joint angular velocity may indicate a weakening of the joint movement trend or be opposite to the increasing trend of the interaction force. Therefore, this increase in human-machine interaction force cannot be directly interpreted as an increased demand for active assistance. To avoid the dimensional differences of different physical quantities affecting the comparison between the protective force enhancement characterization and the subsequent reinjection amplification trend value, the controller performs a directional unification of the human-machine interaction force change rate before calculating the protective force enhancement characterization. Predicted joint angular velocity after unification of direction Perform normalization processing; In one implementation, the controller uses the maximum absolute value of the corresponding physical quantity collected within the current handling cycle as the normalization benchmark; the normalized value of the human-machine interaction force change rate after direction unification and the normalized value of the predicted joint angular velocity after direction unification are respectively: in, This represents the normalized rate of change of human-computer interaction force. This represents the normalized predicted joint angular velocity; This represents the maximum absolute value of the rate of change of human-machine interaction force after the direction is unified up to the current control moment within the current handling cycle; This indicates the maximum absolute value of the predicted joint angular velocity after the direction is unified up to the current control moment within the current handling cycle; This represents an extremely small positive number to prevent the denominator from being zero; the extremely small positive number is determined by the numerical calculation accuracy of the controller. in, and Update the data according to the data collected or calculated within the current handling cycle; when the amount of data before the current control moment is insufficient to form a stable maximum value, the change rate benchmark corresponding to the rated range of the force sensor, the rated range of the joint speed, or the system calibration value can be used as the initial normalization benchmark; during the current handling cycle, the corresponding normalization benchmark is dynamically updated using the collected data. In the implementation method using normalization, the protective force enhancement characteristic quantity Based on the normalized rate of change of human-computer interaction force and normalized predicted joint angular velocities calculate: in, This represents a protective force-enhancing quantity; This represents the normalized rate of change of human-computer interaction force. This represents the normalized predicted joint angular velocity; In the formula This is used to limit the calculation of protective force enhancement parameters only when the human-machine interaction force increases in the positive direction of the target joint assistance; when When the value is not greater than zero, it indicates that the human-computer interaction force after unification and normalization has not increased; at this time, the protective force enhancement quantity is zero or decreases; in the formula Used to determine whether the predicted joint movement trend after direction unification and normalization is opposite to the increasing trend of human-computer interaction force after direction unification and normalization; when and When the product of the two is negative, it indicates that the predicted joint movement trend is opposite to the increasing trend of human-machine interaction force, and the protective force enhancement quantity increases; when the product of the two is not negative, it indicates that the predicted joint movement trend is not opposite to the increasing trend of human-machine interaction force, and the protective force enhancement quantity decreases or becomes zero. Through the above calculation method, the controller does not only judge that the wearer needs stronger assistance based on the increase in human-machine interaction force, but also judges whether the increase in human-machine interaction force is opposite to the predicted joint movement trend under the same target joint assistance direction reference and normalized numerical reference, thereby identifying protective force increase caused by cargo jamming release, center of gravity shift, partial unloading or load shaking.

[0016] S4. Calculation of the rate of change of output torque of exoskeleton joints, the value of recirculation amplification trend, and the increment of recirculation amplification risk; Specifically, the controller is based on the protective force enhancement characteristic quantity. Normalized rate of change of human-computer interaction force and exoskeleton joint output torque The risk of amplification is obtained through recharge. ; The controller outputs torque to the exoskeleton joints based on the current control moment. Exoskeleton joint output torque at the previous sampling time and sampling time interval The rate of change of output torque of the exoskeleton joint was obtained. : in, This represents the rate of change of the exoskeleton joint output torque at the current control moment; This indicates the exoskeleton joint output torque at the current control moment; This represents the exoskeleton joint output torque at the previous sampling time. This represents the sampling time interval between two adjacent sampling moments; this formula is used to characterize the degree of change in the assistive output of the exoskeleton robot; when When positive, it indicates that the exoskeleton joint output torque is increasing; when... When the value is zero or negative, it indicates that the output torque of the exoskeleton joint has not increased or is decreasing; To avoid dimensional differences between the rate of change of exoskeleton joint output torque and other physical quantities, the controller controls the rate of change of exoskeleton joint output torque. Normalization is performed: in, This represents the normalized rate of change of the output torque of the exoskeleton joints. This represents the rate of change of the output torque of the exoskeleton joints; This represents the maximum absolute value of the rate of change of the exoskeleton joint output torque up to the current control time within the current handling cycle; This represents an extremely small positive number to prevent the denominator from being zero; the extremely small positive number is determined by the numerical calculation accuracy of the controller. in, Update based on the data collected or calculated within the current handling cycle; when the amount of data before the current control moment is insufficient to form a stable maximum value, the range of the joint torque sensor, the range of the rated torque variation of the exoskeleton joint, or the system calibration value can be used as the initial normalization reference; during the current handling cycle, the collected data is used to dynamically update the corresponding normalization reference; Through the above normalization process , and All are dimensionless or relative quantities, which ensures that the calculation of protective force enhancement characteristics and reinjection amplification trend values ​​can remain comparable under different joints, different load weights and different sensor ranges.

[0017] The controller is based on the protective force enhancement quantity. Normalized rate of change of human-computer interaction force and normalized rate of change of exoskeleton joint output torque The current control moment's backfeed amplification trend value R(t) is obtained: in, This indicates the current control moment's reinjection amplification trend value; This represents a protective force-enhancing quantity; This represents the normalized rate of change of human-computer interaction force. This represents the normalized rate of change of the output torque of the exoskeleton joints. When the protective enhancement characterization quantity This indicates the presence of protective enhancement, and and When both indicate an increasing trend, the increase in the reinjection amplification trend value R(t) at the current control moment indicates that the increase in human-machine interaction force and the increase in exoskeleton joint output torque occur simultaneously. There is a risk that the exoskeleton assist gain will continue to amplify protective force along the same force chain; when the protective force amplification characteristic quantity... It did not indicate the presence of protective enhancement, or and When the non-uniformity indicates an increasing trend, the reinjection amplification trend value R(t) at the current control time decreases or becomes zero; To avoid poor adaptability to different wearers and different load weights due to the use of a fixed threshold, this embodiment uses the reinjection amplification trend value at the current control moment. Compared with the historical average of the amplification trend during the current transport cycle By comparison, the incremental risk of reinjection amplification is obtained. : in, This indicates that the reinjection amplifies the increased risk at the current control point. This indicates the current control moment's reinjection amplification trend value; This represents the average historical backflow amplification trend within the current transport cycle; Historical rebound amplification trend mean It can be obtained as follows: Where n represents the number of historical control moments sampled within the current transport cycle; This represents the i-th historical control moment within the current transport cycle; This represents the reinjection amplification trend value at the i-th historical control moment; when When the value is greater than 0, it indicates that the current control moment's backfeed amplification trend value is higher than the historical average level within the current handling cycle, and the controller determines that the current control moment is in a state of increased backfeed amplification risk; when When the current control moment indicates that the backflow amplification trend value is not higher than the historical average level within the current transport cycle, the controller determines that the current control moment is not in a state of increased backflow amplification risk.

[0018] S5. Normal expected assist torque, risk suppression coefficient and control torque generation; Specifically, the controller obtains the control torque based on the predicted joint motion state, the target joint motion state, and the incremental risk of recirculation amplification. ; The controller predicts the joint angle. Angle with the target joint The angle difference between them, predicting the joint angular velocity Angular velocity of the target joint The angular velocity difference between them, along with impedance control parameters including the impedance stiffness coefficient and damping coefficient, yields the normal expected assist torque. : in, K(t) represents the normal expected assist torque at the current control moment; K(t) represents the impedance stiffness coefficient at the current control moment. This represents the damping coefficient at the current control moment; This indicates the predicted joint angle; Indicates the target joint angle; This indicates the predicted joint angular velocity; This represents the target joint angular velocity. This formula is a common form in impedance control, meaning that when there is an angular difference between the predicted joint angle and the current target joint angle, an assist torque component corresponding to the angular difference is generated through the impedance stiffness coefficient K(t); when there is an angular velocity difference between the predicted joint angular velocity and the current target joint angular velocity, an assist torque component corresponding to the angular difference is generated through the damping coefficient. Generate the assist torque component corresponding to the angular velocity difference; normal expected assist torque Assistance used to match the output of exoskeleton joints with the predicted joint motion state; The controller amplifies the risk increment based on the recharge. Normal expected assist torque By applying constraints, the control torque can be obtained. In one implementation, the control torque Calculate as follows: in, This indicates the control torque at the current control moment; This represents the normal expected assist torque at the current control moment; This indicates that the reinjection amplifies the increased risk at the current control point. Indicates the risk inhibition coefficient; Among them, the risk inhibition coefficient Used to amplify the risk increment of recharge Converted to normal expected assist torque Constraint strength; Risk suppression coefficient This can be determined through system calibration. During calibration, the controller collects the incremental risk of backflow amplification under no-load handling conditions, stable load handling conditions, and disturbed load handling conditions. The no-load handling condition is used to obtain the baseline of the incremental risk of backflow amplification under the wearer's natural movement state. The stable load handling condition is used to obtain the range of variation of the incremental risk of backflow amplification under normal active handling actions. The disturbed load handling condition is used to obtain the upper limit of the incremental risk of backflow amplification under conditions of cargo jamming release, partial unloading, or center of gravity shift. The controller records the maximum reinjection amplification risk increment obtained under disturbance load handling conditions as . The desired control torque compression ratio is denoted as ,in This indicates the retention ratio of the control torque relative to the normal expected assist torque under the maximum amplified risk increment of backflow, and Risk inhibition coefficient Determine as follows: in, Indicates the risk inhibition coefficient; Indicates the desired control torque compression ratio; This represents the maximum amplified risk increment obtained under the condition of disturbed load handling; In the above way, when equal At that time, control torque Compared to the normal expected assist torque The proportion is approximately This allows for a clear definition of the assist compression level under maximum disturbance risk; the desired control torque compression ratio. The parameters can be determined based on the exoskeleton's rated torque, maximum permissible rate of change of assist, and the wearer's safety and comfort requirements; for example, in heavy-duty transport exoskeletons, It can be set to a ratio that allows the control torque to maintain attitude support without further amplifying strong assist; if no disturbance load handling test was performed during the calibration phase, it can also be set to... Set to the maximum allowable increment of backflow amplification risk in the system security calibration; when When the value is zero or close to zero, it indicates that no effective backflow amplification risk has occurred during the calibration process. The controller adopts the system default risk suppression coefficient, which is determined by the rated torque of the exoskeleton joint and the maximum allowable assist change rate, so that the change in control torque in any control cycle does not exceed the maximum allowable assist change. when hour, Zero, control torque Equal to normal expected assist torque This indicates that the current control state is not in a state of heightened risk of backflow amplification, and the exoskeleton robot outputs power according to normal impedance; when When the value is greater than 0, the controller determines that the current control moment is in a state of increased risk of backfeed amplification, and the control torque... Less than the expected assist torque The output amplitude represents the output amplitude of the exoskeleton robot reducing the normal expected assist torque, thereby preventing the exoskeleton assist from continuing to amplify along the direction of increasing human-machine interaction force.

[0019] S6, Assisted Execution, Gradual Recovery and Safety Correction; Specifically, the controller will control the torque. Send to the exoskeleton joint actuator to make the exoskeleton joint follow the control torque Output assistance; the joint actuator adjusts according to the control torque. Adjust the motor output so that the exoskeleton joints provide auxiliary torque to the wearer's target joints; At the next control moment, the controller reacquires the human-machine collaboration status data and repeats S1 to S5 to obtain the incremental risk of backfeed amplification at the next control moment and the normal expected assist torque at the next control moment. In order to avoid prematurely restoring high assist output before the risk of backfeed amplification has been eliminated, the controller performs a safety correction at the next control moment. In one implementation, the controller calculates the safety closed-loop change. : in, This represents the change in the safety closed-loop control at the current moment. This represents the human-computer interaction force after the direction is unified at the current control moment; This represents the human-computer interaction force after the direction is unified at the previous sampling time. This indicates the control torque at the current control moment; This indicates the control torque at the previous sampling time. This represents the proportionality coefficient, used to unify the degree of influence of changes in human-machine interaction force and control torque on the change in the safety closed loop. Safety closed-loop change This is used to reflect whether changes in human-machine interaction force and control torque tend to stabilize; when both changes in human-machine interaction force and control torque decrease, the change in the safety closed-loop is... The decrease indicates that the reinjection amplification chain is being suppressed; when the changes in human-machine interaction force and control torque do not decrease, the change in the safety closed loop is... The fact that it did not decrease indicates that there may still be a risk of instability in the human-machine collaboration state; Specifically, when the incremental risk of backflow amplification decreases or is no longer in a state of increased backflow amplification risk at the next control moment, the controller uses the normal expected assist torque at the next control moment as the control torque for the next control moment, so that the exoskeleton robot can restore normal impedance assist. To avoid the wearer feeling an impact due to the control torque jumping directly from the constrained state to the normal expected assist torque, the controller adopts a gradual recovery method when restoring normal impedance assist. Specifically, when the incremental risk of backflow amplification decreases or is no longer in a state of increased backflow amplification risk at the next control moment, the controller first calculates the normal expected assist torque at the next control moment and compares the difference between the normal expected assist torque at the next control moment and the control torque at the previous control moment. If the absolute value of the difference is not greater than the maximum allowable assist change, the normal expected assist torque at the next control moment is used as the control torque for the next control moment. If the absolute value of the difference is greater than the maximum allowable assist change, the control torque is gradually increased or decreased according to the maximum allowable assist change until the control torque reaches the normal expected assist torque. The maximum permissible change in assist can be determined by the exoskeleton joint rated torque, the maximum torque change rate of the joint actuator, and the control cycle. For example, the maximum permissible change in assist can be the product of the maximum torque change rate of the joint actuator and the control cycle, or it can be the maximum single-cycle torque change that will not cause significant impact to the wearer, as determined by the exoskeleton system calibration. If the risk increment of backflow amplification is greater than zero again during the progressive recovery process, the controller stops the recovery process and re-constrains the normal expected assist torque according to the risk increment of backflow amplification. When the risk increment of backfeed amplification continues to be in a state of enhanced backfeed amplification risk at the next control moment, the controller continues to reduce the output amplitude of the normal expected assist torque at the next control moment, so that the exoskeleton robot maintains constrained assist output. When the human-machine interaction force increases during continuous control moments and the change in control torque does not decrease, the controller reduces the assist output or triggers a safety shutdown protection. The increase in human-machine interaction force during continuous control moments means that the human-machine interaction force after the direction of the subsequent control moment is greater than the human-machine interaction force after the direction of the preceding control moment is unified at least twice in consecutive control moments. The lack of decrease in the change in control torque means that the change in control torque at the current control moment is not less than the change in control torque at the preceding control moment. The change in control torque at the current control moment is the absolute value of the difference between the control torque at the current control moment and the control torque at the preceding control moment.

[0020] If the human-machine interaction force increases continuously and the change in control torque does not decrease, and if the human-machine interaction force after unification of direction does not exceed the safety limit of the human-machine interaction force allowed by the exoskeleton system, then the controller reduces the assist output. When reducing the assist output, the controller reduces the current control torque to the posture support torque according to a preset reduction slope. The posture support torque is the torque that maintains the joint support posture of the exoskeleton without further increasing the handling assistance. The preset reduction slope is determined by the maximum allowable torque change rate of the joint actuator and the control cycle. If, under the condition that the human-machine interaction force continuously increases and the change in control torque does not decrease, the human-machine interaction force after unification of direction reaches or exceeds the safety limit of the human-machine interaction force allowed by the exoskeleton system, or the output torque of the exoskeleton joint reaches or exceeds the rated torque of the exoskeleton joint, the controller triggers a safety shutdown protection. The safety shutdown protection includes stopping the target joint actuator from continuing to output active assistance and putting the exoskeleton joint into a braking, limiting, or passive support state to prevent the exoskeleton robot from continuing to output assistance in the direction of increasing human-machine interaction force. The safety limit of human-machine interaction force can be determined during the system calibration stage based on the allowable contact force of the contact part between the exoskeleton robot and the human body, the load-bearing capacity of the straps, and the safety and comfort requirements of the wearer. The rated torque of the exoskeleton joint is determined by the rated parameters of the exoskeleton joint actuator and the transmission mechanism. Through the above correction process, the exoskeleton robot can smoothly restore normal assistance when the risk of backflow amplification decreases, limit assistance output when the risk of backflow amplification continues to increase, and implement safety protection when the human-machine interaction state is continuously unstable, thereby avoiding the continuous amplification of load disturbance, human protective force increase and exoskeleton assistance gain along the same force chain.

[0021] In another embodiment, if the exoskeleton robot includes multiple target joints, the controller can calculate the predicted joint angle, predicted joint angular velocity, protective force enhancement quantity, refeedback amplification risk increment, and control torque for each target joint. When any one of the multiple target joints is in a state of enhanced refeedback amplification risk, the controller can only constrain the normal expected assist torque of that target joint. When multiple target joints are simultaneously in a state of enhanced refeedback amplification risk, the controller can simultaneously constrain the normal expected assist torque of multiple target joints to prevent changes in local joint assistance from causing overall posture imbalance of the wearer. Specifically, in a multi-joint exoskeleton robot, the target joint may include multiple exoskeleton-assisted joints; the controller numbers the multiple target joints as the j-th target joint, where j is the target joint number; for the j-th target joint, the controller collects the target joint angles respectively. Target joint angular velocity Target joint angular acceleration Human-computer interaction and exoskeleton joint output torque And calculate the predicted joint angles respectively. Predicting joint angular velocity Human-computer interaction power after unification of direction Rate of change of human-computer interaction force after unification of direction Normalized rate of change of human-computer interaction force Normalized predicted joint angular velocities Normalized rate of change of exoskeleton joint output torque Protective force enhancement characterization quantity Reinjection amplifies the risk increment and control torque ; When the reperfusion amplification risk increment of the j-th target joint When the target joint is in a state of increased risk of recirculation amplification, the controller reduces the normal expected assist torque output amplitude of the j-th target joint to obtain the control torque of the target joint. When multiple target joints are simultaneously in a state of heightened risk of backflow amplification, the controller constrains the normal expected assist torque of the corresponding target joints. If one target joint is in a state of heightened risk of backflow amplification while the adjacent target joints are not, the controller only constrains the target joint in the state of heightened risk of backflow amplification and maintains the adjacent target joints to output assist according to the corresponding control torque, thereby avoiding the excessive reduction of assist by multiple joints due to local load disturbances, which would affect the wearer's posture support. During multi-joint control, if the human-machine interaction force of any target joint reaches or exceeds the corresponding human-machine interaction force safety limit, or if the exoskeleton joint output torque of any target joint reaches or exceeds the corresponding exoskeleton joint rated torque, the controller will trigger safety protection for that target joint; if multiple target joints trigger safety protection conditions simultaneously, the controller will synchronously execute safety protection for multiple target joints.

[0022] In another implementation, the historical recharge amplification trend mean The arithmetic mean of the current handling cycle, the sliding window mean, or the exponential moving average can be used. When using the sliding window mean, the controller only uses the feedback amplification trend values ​​of multiple consecutive control times before the current control time to calculate the historical feedback amplification trend mean, so as to improve the response speed to short-term load disturbances. When using the exponential moving average, the feedback amplification trend values ​​of more recent control times have higher weights so that the controller can more sensitively reflect the current handling status. Regardless of the method used, the historical feedback amplification trend mean is derived from the data that has been collected and calculated within the current handling cycle.

[0023] In another embodiment, the normal expected assist torque It can be obtained from the impedance control relationship, or the amplitude can be limited by combining the rated torque limit of the exoskeleton joint, the maximum output capability of the driver, or the mechanical limit of the joint on the basis of impedance control; the above amplitude limit is only used to ensure the safety of the exoskeleton hardware and does not change the core logic of the present invention to assist the output by amplifying the risk increment constraint through refeedback.

[0024] In another embodiment, the exoskeleton joint outputs torque. Instead of being directly collected by the joint torque sensor, the torque can be calculated from the driver current. Specifically, the controller obtains the exoskeleton joint output torque based on the driver current, motor torque constant, reduction gear transmission ratio, and transmission efficiency. This calculation method is a conventional torque estimation method in motor drive systems and can provide information on the change in exoskeleton joint output torque for calculating the risk increment of the reinjection amplification.

[0025] In another embodiment, the controller can measure the rate of change of the human-machine interaction force after directional unification before calculating the protective force enhancement quantity. Predicted joint angular velocity after direction unification and the rate of change of output torque of exoskeleton joints Filtering is performed to reduce the impact of sensor noise. Filtering can be done using moving average filtering or low-pass filtering. The filtered data still represents the trend of human-machine interaction force changes, predicted joint movement trends, and exoskeleton assist output changes near the corresponding control time. It does not change the role of the protective force enhancement quantity in identifying whether the increase in interaction force is opposite to the predicted joint movement trend, nor does it change the role of the refeedback amplification risk increment in judging whether the exoskeleton assist is amplified along the same force chain.

[0026] To further illustrate the technical effects of this embodiment, a comparative example is set as an exoskeleton robot control method that adjusts assistance solely based on the magnitude of human-machine interaction force. In the comparative example, when the human-machine interaction force increases, the controller directly increases the output torque of the exoskeleton joints or increases the impedance stiffness. This method can provide strong assistance when the wearer actively lifts the load, but when the load is disturbed due to cargo jamming, center of gravity shift, or partial unloading, the protective force generated by the wearer will also lead to an increase in the human-machine interaction force. The comparative example cannot distinguish whether the increase in human-machine interaction force is due to an increased demand for active assistance or a protective force increase, so it may continue to increase the assistance output, causing the interaction force to increase further, resulting in assistance overshoot or human-machine conflict. Compared to the comparative example, this embodiment does not directly equate an increase in human-machine interaction force with an increase in active assistance demand. Instead, it determines the joint movement trend by predicting the joint angular velocity and determines the increasing trend of the interaction force by using the normalized and directional human-machine interaction force change rate. When the increasing trend of the interaction force is opposite to the predicted joint movement trend, a protective force enhancement characteristic is calculated. When the protective force enhancement characteristic exists and the exoskeleton joint output torque is also increasing, the refeedback amplification risk increment is calculated. When the refeedback amplification risk increment indicates an increased refeedback amplification risk, the output amplitude of the normal expected assistance torque is reduced. Thus, this embodiment can block the refeedback amplification chain between load disturbance, human protective force enhancement, and exoskeleton assistance gain, improving the compliance and safety of the industrial handling exoskeleton robot under complex load disturbance conditions.

[0027] This embodiment collects the target joint motion state, human-machine interaction force, and exoskeleton joint output torque to construct a continuous calculation link between the predicted joint motion state, the rate of change of human-machine interaction force after direction unification, the protective force amplification characterization quantity, the refeedback amplification trend value, and the refeedback amplification risk increment. This enables the exoskeleton robot to identify whether the increase in interaction force has protective force amplification characteristics and further determine whether the exoskeleton assistance output is generating refeedback amplification along the direction of the increase in interaction force. This embodiment uses directional unification processing to enable comparison of predicted joint angular velocity and human-machine interaction force change rate under the same target joint assistance positive direction; normalization processing makes motion quantities, force changes, and torque changes of different dimensions comparable; risk suppression coefficient calibration makes the incremental risk of backflow amplification clearly convertible into control torque constraint strength; and progressive recovery and safety closed-loop correction enable the exoskeleton robot to smoothly restore normal assistance after the risk of backflow amplification decreases, continue to limit assistance output when the risk continues to increase, and reduce assistance or trigger safety shutdown protection when the human-machine interaction state remains unstable. Therefore, this embodiment can avoid misinterpreting human protective force enhancement as active assistance demand enhancement in industrial handling load disturbance scenarios, reduce the risk of assistance overshoot, human-machine confrontation and handling posture instability, and improve the stability, safety and compliance of the exoskeleton robot's adaptive collaborative control.

[0028] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An adaptive cooperative control method for exoskeleton robots, characterized in that, Includes the following steps: S1. Collect human-machine collaboration status data within the current handling cycle to obtain the target joint motion status, human-machine interaction force, and exoskeleton joint output torque; S2. Obtain the predicted joint motion state based on the target joint motion state, and obtain the human-computer interaction force change rate based on the human-computer interaction force; S3. Obtain the protective force enhancement characterization quantity based on the predicted joint motion state and the human-machine interaction force change rate; S4. Obtain the reinjection amplification risk increment based on the protective force enhancement characterization quantity, the human-machine interaction force change rate, and the exoskeleton joint output torque; S5. The control torque is obtained based on the predicted joint motion state, the target joint motion state, and the reperfusion amplification risk increment. S6. Drive the exoskeleton joint to output assistance according to the control torque, and perform safety closed-loop correction based on the human-machine collaboration status data at the next control moment.

2. The adaptive cooperative control method for exoskeleton robots according to claim 1, characterized in that, The human-machine collaboration status data includes target joint angle, target joint angular velocity, target joint angular acceleration, human-machine interaction force, and exoskeleton joint output torque. The target joint motion state includes target joint angle, target joint angular velocity, and target joint angular acceleration. The target joint corresponding to the target joint motion state is an exoskeleton-assisted joint that participates in lifting, lowering, or transporting actions during industrial handling.

3. The adaptive cooperative control method for exoskeleton robots according to claim 2, characterized in that, The collection of human-machine collaboration status data within the current handling cycle includes: Acquire the target joint angle output by the inertial measurement unit or joint encoder; The target joint angular velocity and target joint angular acceleration are obtained by an inertial measurement unit or by processing the target joint angle over time. Collect the human-machine interaction force output from force sensors installed at human-machine interface or joint transmission parts; The exoskeleton joint output torque is collected from the joint torque sensor, or the exoskeleton joint output torque is calculated based on the driver current and joint transmission parameters. The target joint angle, target joint angular velocity, target joint angular acceleration, human-computer interaction force, and exoskeleton joint output torque are synchronized according to the sampling time.

4. The adaptive cooperative control method for exoskeleton robots according to claim 2, characterized in that, The predicted joint motion state is obtained based on the target joint motion state, including: Based on the target joint angle, target joint angular velocity and target joint angular acceleration at the current control moment, joint motion extrapolation is performed within a short-time prediction window to obtain the predicted joint angle and predicted joint angular velocity. The predicted joint motion state includes the predicted joint angle and the predicted joint angular velocity.

5. The adaptive cooperative control method for exoskeleton robots according to claim 1, characterized in that, The rate of change of the human-computer interaction force is obtained based on the human-computer interaction force, including: Obtain the human-computer interaction force at the current control moment and the human-computer interaction force at the previous sampling moment; The rate of change of the human-computer interaction force is obtained based on the human-computer interaction force at the current control moment, the human-computer interaction force at the previous sampling moment, and the sampling time interval between two adjacent sampling moments.

6. The adaptive cooperative control method for exoskeleton robots according to claim 1, characterized in that, The protective force enhancement characterization quantity is obtained based on the predicted joint motion state and the human-machine interaction force change rate, including: When the rate of change of human-computer interaction force indicates an increase in human-computer interaction force, and the trend of joint movement represented by the predicted joint angular velocity is opposite to the trend of the increase in human-computer interaction force, the protective force enhancement characterization quantity is increased. When the rate of change of human-computer interaction force indicates that the human-computer interaction force has not increased, or when the joint motion trend represented by the predicted joint angular velocity is not opposite to the increasing trend of the human-computer interaction force, the protective force enhancement characterization quantity is reduced.

7. The adaptive cooperative control method for exoskeleton robots according to claim 1, characterized in that, The incremental risk of re-injection amplification is obtained based on the protective force enhancement characteristic, the rate of change of human-machine interaction force, and the output torque of the exoskeleton joint, including: The rate of change of exoskeleton joint output torque is obtained based on the exoskeleton joint output torque at the current control moment, the exoskeleton joint output torque at the previous sampling moment, and the sampling time interval between two adjacent sampling moments. Based on the protective force enhancement characteristic quantity, the human-machine interaction force change rate, and the exoskeleton joint output torque change rate, the reinjection amplification trend value at the current control moment is obtained; The reinjection amplification trend value at the current control moment is compared with the historical reinjection amplification trend average within the current transport cycle to obtain the reinjection amplification risk increment.

8. The adaptive cooperative control method for exoskeleton robots according to claim 7, characterized in that, Based on the protective force enhancement characteristic quantity, the rate of change of human-machine interaction force, and the rate of change of output torque of the exoskeleton joint, the reinjection amplification trend value at the current control moment is obtained, including: When the protective force enhancement characteristic indicates the existence of protective force enhancement, and both the rate of change of human-machine interaction force and the rate of change of output torque of exoskeleton joints show an increasing trend, the reinjection amplification trend value at the current control moment is increased. When the protective force enhancement characteristic does not indicate the existence of protective force enhancement, or when the rate of change of human-machine interaction force and the rate of change of output torque of exoskeleton joint do not both indicate an increasing trend, the recharge amplification trend value at the current control moment is reduced.

9. The adaptive cooperative control method for exoskeleton robots according to claim 7, characterized in that, The control torque is obtained based on the predicted joint motion state, the target joint motion state, and the amplified risk increment of the reperfusion, including: Based on the angle difference between the predicted joint angle and the target joint angle, the angular velocity difference between the predicted joint angular velocity and the target joint angular velocity, and the impedance control parameters including the impedance stiffness coefficient and the damping coefficient, the normal expected assist torque is obtained. When the incremental risk of reinjection amplification indicates that the current control moment is in a state of enhanced reinjection amplification risk, the output amplitude of the normal expected assist torque is reduced to obtain the control torque; When the incremental risk of reinjection amplification indicates that the current control moment is not in a state of enhanced reinjection amplification risk, the normal expected assist torque is used as the control torque.

10. The adaptive cooperative control method for exoskeleton robots according to claim 1, characterized in that, Drive the exoskeleton joints to output assistance according to the control torque, and perform safety corrections based on the human-machine collaboration state data at the next control moment, including: The control torque is sent to the exoskeleton joint actuator, so that the exoskeleton joint outputs assistance according to the control torque; The human-machine collaboration status data is reacquired at the next control moment, and the risk increment amplified by the feedback at the next control moment is obtained. When the incremental risk of reinjection amplification decreases or is not in a state of increased reinjection amplification risk at the next control moment, the normal expected assist torque at the next control moment shall be used as the control torque at the next control moment. When the risk increment of the backfeed amplification at the next control moment continues to be in a state of enhanced backfeed amplification risk, the output amplitude of the normal expected assist torque at the next control moment will be further reduced. When the human-machine interaction force increases during continuous control and the change in control torque does not decrease, reduce the assist output or trigger the safety shutdown protection.

Images