Actuator force fighting regulation and control method and device based on adaptive state fusion

Abstract

The invention discloses an actuator force fighting regulation and control method and device based on adaptive state fusion, and relates to the field of flight control, and the method comprises the steps: collecting the state data of a plurality of redundant actuators in real time; generating a static pressure difference residual error, a dynamic response difference, a displacement difference, an output force sudden change and a state instability index, and calculating a dynamic difference characteristic; the fusion weight of each actuator is calculated through fuzzy control, a neural network or rule logic dynamics; generating an independent compensation amount of each channel by generating a bias compensation instruction; correcting the main control instruction to generate a final actuator instruction; and when the dynamic residual error of a certain actuator is abnormal or the health index is lower than a threshold value, fault stripping processing is carried out. The method has the advantages that real-time suppression of internal force conflicts and smooth stripping of fault channels are achieved through multi-dimensional state fusion and self-adaptive weight distribution, and the effects of improving control surface control precision, reducing structural loss and guaranteeing fault tolerance are achieved.

Description

Technical Field

[0001] The present invention relates to the field of flight control, and in particular to a method and device for controlling actuator force conflicts based on adaptive state fusion. Background Art

[0002] With the increasing demands for flight control safety and stability in new-generation aircraft such as large commercial aircraft and eVTOLs, flight control actuation systems generally adopt redundant parallel actuator structures. However, due to differences in manufacturing, installation, aging, and controller delays among redundant channels, multiple actuators can experience "force contention" when driving the same control surface. This means that the actuators restrain each other, causing structural fatigue, energy loss, and reduced control accuracy.

[0003] Most existing methods are based on simple threshold logic such as static force difference and displacement difference for monitoring, or use pressure difference balancing control to achieve load distribution. As a result, they are not able to effectively deal with the "asynchronous characteristics" such as response speed and stabilization time between channels under dynamic operation conditions during use. In addition, when the parameters are fixed, they cannot adaptively adjust to environmental changes such as actuator health status and temperature. At the same time, most of them only monitor or alleviate, but do not have self-learning / dynamic compensation capabilities. Summary of the Invention

[0004] In order to solve the above technical problems, a method and device for controlling actuator force disputes based on adaptive state fusion are provided. This technical solution solves the problem of force disputes caused by dynamic response differences of redundant actuators proposed in the above background technology.

[0005] In order to achieve the above objects, the technical solution adopted by the present invention is:

[0006] A method and device for controlling actuator force contention based on adaptive state fusion, comprising:

[0007] Real-time collection of status data of multiple redundant actuators, including output displacement, output force or pressure difference, response delay, status noise and health indicators;

[0008] Based on the reference trajectory and the state data of each actuator, the static pressure difference residual, dynamic response difference, displacement difference, output force mutation and state instability index are generated, and the dynamic difference characteristics are calculated;

[0009] According to the dynamic difference characteristics and actuator health indicators, the fusion weight of each actuator is calculated through fuzzy control, neural network or rule logic dynamics;

[0010] Based on the displacement difference, static pressure difference residual and weight, the offset compensation instruction is generated according to Δcmd i =K i ×(ΔX i +ΔPi ) generates independent compensation for each channel, where K i is the dynamic adjustment coefficient;

[0011] Superimpose the offset command on the flight control main command, correct the main control command, and generate the final actuator command;

[0012] When the dynamic residual of an actuator is abnormal or the health index is lower than the threshold, fault separation is performed, its weight is reduced and it is prohibited from participating in the control output.

[0013] Preferably, the dynamic difference features include:

[0014] Static pressure difference residual: the steady-state deviation between the actual output force or pressure difference of each actuator and the target value;

[0015] Dynamic response difference: The actuator response delay time difference estimated by step response;

[0016] Displacement difference: The deviation between the actual displacement of the actuator and the reference trajectory, which includes the average displacement or target velocity trajectory;

[0017] Output force mutation: the instantaneous rate of change of the actuator output force;

[0018] State instability index: The statistical value of the actuator state noise calculated based on the signal fluctuation rate.

[0019] Preferably, the fusion weight is calculated in at least one of the following ways:

[0020] Fuzzy control: output weights based on the fuzzy rule base of dynamic difference characteristics and health indicators;

[0021] Rule logic: Dynamically assign weights through preset threshold logic or empirical rules;

[0022] Preferably, the generation formula of the offset compensation instruction is Δcmd i =K i ×(ΔX i +ΔP i ), where K i The coefficient of real-time adjustment is obtained through fuzzy control or neural network dynamic optimization based on dynamic difference characteristics, ΔX i is the current channel displacement difference, ΔP i is the static pressure difference residual.

[0023] Preferably, the fault isolation process includes:

[0024] When the dynamic residual continues to exceed the set threshold or the health indicator is lower than the safety lower limit;

[0025] Reduce the weight of the actuator to zero and prohibit it from participating in the control output;

[0026] Perform soft privilege reduction or complete isolation operations to ensure system fault tolerance.

[0027] Preferably, the health indicator is derived through built-in self-test or cumulative performance degradation data, wherein the built-in self-test is a real-time diagnosis of the electrical or mechanical performance of the actuator, and the cumulative performance degradation data is a health score based on historical status data, and the historical status data includes response delay and noise growth.

[0028] Preferably, the reference trajectory is an average value of displacement feedback of each actuator or a target speed trajectory issued by a flight control system.

[0029] A device for implementing the method according to any one of claims 1 to 7, comprising:

[0030] State acquisition module: acquires the actuator's output displacement, output force or pressure difference, response delay, state noise and health indicators in real time;

[0031] Difference analysis module: calculates the dynamic difference characteristics of each actuator relative to the reference trajectory;

[0032] Adaptive weight allocator: Based on dynamic difference characteristics and health indicators, it generates fusion weights through fuzzy control and neural network;

[0033] Bias generation module: calculates bias instructions based on displacement difference, static pressure difference residual and weight;

[0034] Main instruction correction module: superimposes the offset instruction on the flight control main instruction to generate the final actuator instruction;

[0035] Fault isolation module: performs privilege reduction and output isolation operations on abnormal channels.

[0036] Preferably, the dynamic difference characteristics output by the difference analysis module include static pressure difference residual, dynamic response difference, displacement difference, output force mutation and state instability index.

[0037] Preferably, the bias generation module performs the operation Δcmd i =K i ×(ΔX i +ΔP i ), where K i Dynamically adjusted by an adaptive weight allocator.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] The present invention proposes a method and device for controlling actuator force contention based on adaptive state fusion. By injecting step commands, the delay difference from 10% to 90% response of each channel is measured, dynamic asynchrony is quantified, and dynamic response difference is calculated. The force signal variance is calculated in real time using a 100ms sliding window, and noise growth is captured for state instability analysis. The mean trajectory (balanced load) or target trajectory (precise tracking) is automatically selected according to the flight control mode. This method can simultaneously cover steady-state deviation, transient response, and noise statistical characteristics, overcoming the limitation of the static threshold method that cannot handle dynamic asynchrony (such as increased delay due to low temperature). Simulations also show that displacement tracking error is reduced and the peak value of internal force conflict is reduced.

[0040] The present invention proposes a method and device for controlling actuator force disputes based on adaptive state fusion. By including operating conditions such as temperature and aging in training data, the weight distribution is adapted to nonlinear disturbances (such as transonic aerodynamic flutter) and a dual-loop collaborative mechanism including a compensation loop and a fault-tolerant loop. This allows the compensation command to act directly on the rudder position and the internal force balance point, which reduces energy consumption and increases the effectiveness of dynamic compensation compared to the traditional pressure difference balancing method. The soft de-weighting strategy can also be used to reduce the switching impact force fluctuation, ensure the continuous operation of the redundant system, and achieve seamless fault separation. At the same time, the neural network weight mapping model maintains control stability in the full temperature range of -40°C to 85°C, which can improve the adaptive generalization capability. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of the process of the present invention;

[0042] Figure 2 It is a schematic diagram of the process of the framework structure in the present invention;

[0043] Figure 3 This is a control flow chart of the present invention;

[0044] Figure 4 Generate a logic diagram for the bias command of the present invention;

[0045] Figure 5 This is a simulation test comparison curve in the present invention. DETAILED DESCRIPTION

[0046] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0047] Reference Figure 1 As shown, a method and device for controlling actuator force contention based on adaptive state fusion includes:

[0048] Real-time collection of status data of multiple redundant actuators, including output displacement, output force or pressure difference, response delay, status noise and health indicators;

[0049] Based on the reference trajectory and the state data of each actuator, the static pressure difference residual, dynamic response difference, displacement difference, output force mutation and state instability index are generated, and the dynamic difference characteristics are calculated;

[0050] According to the dynamic difference characteristics and actuator health indicators, the fusion weight of each actuator is calculated through fuzzy control, neural network or rule logic dynamics;

[0051] Based on the displacement difference, static pressure difference residual and weight, the offset compensation instruction is generated according to Δcmd i =K i ×(ΔX i +ΔP i ) generates independent compensation for each channel, where K i is the dynamic adjustment coefficient;

[0052] Superimpose the offset command on the flight control main command, correct the main control command, and generate the final actuator command;

[0053] When the dynamic residual of an actuator is abnormal or the health index is lower than the threshold, fault separation is performed, its weight is reduced and it is prohibited from participating in the control output.

[0054] The output displacement, force / pressure difference, response delay (estimated by step response), state noise (signal fluctuation rate) and health indicators (built-in self-test or historical attenuation data derivation) of redundant actuators are collected in real time through multi-source sensors. Based on the reference trajectory (such as the average displacement or flight control target speed trajectory), the dynamic difference characteristics of each channel are calculated, including the static pressure difference residual (steady-state force deviation), dynamic response difference (time delay difference), displacement difference, output force instantaneous mutation rate and state instability index. Fuzzy control, neural network or preset rule logic are used to dynamically generate the fusion weight of each actuator in combination with health indicators. According to the formula Δcmd i =K i ×(ΔX i +ΔP i )Generate independent offset compensation instructions (K i is a dynamic coefficient) and is added to the flight control main command to implement command correction. When the dynamic residual of a channel is abnormal (continuously exceeding the threshold) or the health score falls below the safety lower limit, soft demotion is performed (the weight is reset to zero) and its output is prohibited.

[0055] During state acquisition, displacement and force signals can be acquired in real time through the built-in sensors of the actuators, and response delay and noise statistics can be calculated in combination with the controller timing data. Health indicators are generated by periodic electrical / mechanical self-tests and historical performance degradation models (such as delay growth and noise aggravation). During differential analysis, the reference trajectory uses the multi-channel displacement average or the target trajectory of the flight control command to calculate the displacement deviation ΔX in real time.i , steady-state pressure difference ΔP i ; Dynamic response difference ΔT i The output force mutation rate is obtained by comparing the step response rise time; the output force mutation rate is obtained by the differential operation of the force signal. The adaptive weight distribution can be obtained by the fuzzy controller with ΔP i , ΔT i , σ i (instability index) is the input, output weight w i ; or the neural network trains the weight mapping model based on historical fault data. At the same time, the bias instruction Δcmd i Dynamically compensate for channel differences; after the faulty channel is removed, it only participates in monitoring and no longer outputs control variables.

[0056] This not only addresses the inability of traditional methods to handle dynamic response variations, but also improves contention identification accuracy through multi-dimensional feature fusion (static, dynamic, and statistical). It also employs an adaptive weighting mechanism to address changes such as actuator aging and environmental disturbances, avoiding misjudgments caused by fixed thresholds (such as increased response delays at low temperatures). Furthermore, a soft weight reduction strategy reduces the risk of mis-switching, ensures system redundancy, and reduces contention peaks through simulation verification.

[0057] The dynamic difference features include:

[0058] Static pressure difference residual: the steady-state deviation between the actual output force or pressure difference of each actuator and the target value;

[0059] Dynamic response difference: The actuator response delay time difference estimated by step response;

[0060] Displacement difference: The deviation between the actual displacement of the actuator and the reference trajectory, which includes the average displacement or target velocity trajectory;

[0061] Output force mutation: the instantaneous rate of change of the actuator output force;

[0062] State instability index: The statistical value of the actuator state noise calculated based on the signal fluctuation rate.

[0063] The static pressure differential residual measures the steady-state deviation between the actuator's actual output force and the target value (e.g., the difference between the pressure sensor reading and the command for a hydraulic actuator). The dynamic response difference, measured by injecting a step command and measuring the time delay between each channel reaching 63% of the target value, is obtained by injecting a step command and measuring the time delay between each channel's response from 10% to 90%. The displacement difference represents the real-time deviation between the actual displacement and the reference trajectory (multi-channel average or flight control target trajectory). Sudden force changes are detected by temporally differentiating the force signal and calculating the instantaneous rate of change to capture abnormal impacts. The state instability indicator, based on a sliding window, measures the variance of signal fluctuations and quantifies the noise level. The variance σ of the force signal within the most recent 100ms window is calculated. This not only addresses the inability of traditional methods to handle dynamic response differences, but also improves contention identification accuracy through multi-dimensional feature fusion (static, dynamic, and statistical). An adaptive weighting mechanism addresses changes such as actuator aging and environmental disturbances, avoiding misjudgments caused by fixed thresholds (e.g., increased response delay at low temperatures). A soft weight reduction strategy reduces the risk of false trips, ensuring system redundancy. Simulations verify the reduction of force contention peaks.

[0064] The fusion weight is calculated in at least one of the following ways:

[0065] Fuzzy control: output weights based on the fuzzy rule base of dynamic difference characteristics and health indicators;

[0066] Rule logic: Dynamically assign weights through preset threshold logic or empirical rules;

[0067] By building a rule base, i.e., a fuzzy controller, the membership function of the input variables (ΔPi, ΔTi, σi) is set, and the weighted output w is obtained through the rule base. i , and then presetting a threshold strategy, the model can be trained using historical data (dynamic features as input, weight values ​​as output). The neural network adopts a three-layer feedforward structure, with dynamic difference features as the input layer, ReLU as the activation function for hidden layer nodes, and Sigmoid normalized weight values ​​in the output layer, supporting online learning. Its fuzzy logic and rule base provide interpretability, facilitating debugging by engineers, while the neural network adapts to complex nonlinear operating conditions (such as transonic aerodynamic disturbances), improving the generalization of weight allocation, and improving load balancing accuracy compared to traditional fixed-gain allocation.

[0068] The generation formula of the offset compensation instruction is Δcmd i =K i ×(ΔX i +ΔP i ), where K i The coefficient of real-time adjustment is obtained through fuzzy control or neural network dynamic optimization based on dynamic difference characteristics, ΔX i is the current channel displacement difference, ΔP i is the static pressure difference residual.

[0069] The bias generation module receives the ΔX from the difference analysis module in real time. i ,ΔP i and K of the weight allocator i , perform linear superposition operation. Generate the formula Δcmd i =K i ×(ΔX i +ΔP i ), where K i Dynamically optimized by the adaptive weight allocator: K i and fusion weight w i Positive correlation (such as K i =αw i ), the health channel obtains higher compensation authority: ΔX i is the current displacement deviation, ΔP i is the static pressure difference residual. Thus, the root cause of force disputes (geometric position deviation + internal force imbalance) can be directly suppressed by jointly compensating the displacement difference and the pressure difference residual, and K i Dynamic adjustment avoids overcompensation and reduces rudder surface offset errors.

[0070] The fault isolation process includes:

[0071] When the dynamic residual continues to exceed the set threshold or the health indicator is lower than the safety lower limit;

[0072] Reduce the weight of the actuator to zero and prohibit it from participating in the control output;

[0073] Perform soft privilege reduction or complete isolation operations to ensure system fault tolerance.

[0074] By continuously exceeding the threshold value (such as ΔT i >50ms) or the health score falls below a safe value (e.g., 80 points); the weight is reset to zero, prohibiting the channel from participating in control output; the weight decays exponentially to zero, avoiding soft weight reduction due to command jumps. A fault stripping process is constructed, where the fault identification module monitors the trend of the difference indicator and combines it with the health score (e.g., a decay model based on cumulative response delay). Once the threshold is triggered, a stripping command is sent to the weight allocator, simultaneously notifying the main command correction module to remove the channel output. Early stripping of the faulty channel can maintain stable pressure differential balance while reducing system impact, resulting in lower force fluctuations than traditional hard switching.

[0075] The health indicator is derived through built-in self-test or accumulated performance degradation data, where the built-in self-test is a real-time diagnosis of the electrical or mechanical performance of the actuator, and the accumulated performance degradation data is a health score based on historical status data, which includes response delay and noise growth.

[0076] By periodically injecting test signals, the built-in self-test (BIST) detects the electrical (coil impedance) and mechanical (friction resistance) performance of the actuator, and constructs a health scoring model based on historical data (such as the response delay growth rate and the rising trend of the noise variance), the cumulative performance degradation is formed to form a health indicator. The BIST module is executed once every 10 seconds, applying a small step instruction, and anomalies are judged by the response speed and waveform distortion. The cumulative degradation model records the average delay of 1,000 consecutive operations. If it exceeds the baseline by 20%, the health score is deducted. In this way, the BIST can capture sudden faults (such as coil short circuit) in real time, and the cumulative model can predict progressive aging and estimate the life extension.

[0077] The reference trajectory is the average value of the displacement feedback of each actuator or the target speed trajectory issued by the flight control system.

[0078] By obtaining the multi-channel displacement average value of all the arithmetic mean values ​​that can be used as actuator displacement feedback, and directly using the flight control target speed trajectory of the ideal control surface motion trajectory issued by the flight control computer to generate two reference trajectories, the difference analysis module receives the actuator displacement data in real time. If the average value mode is used, the calculation If the flight control target trajectory is used, then the instruction X is read directly cmd As a benchmark, the average value mode enhances the self-coordination of the redundant system, while the flight control target trajectory mode is suitable for high-precision tracking scenarios and reduces the control surface trajectory deviation.

[0079] A device for implementing the method according to any one of claims 1 to 7, comprising:

[0080] State acquisition module: acquires the actuator's output displacement, output force or pressure difference, response delay, state noise and health indicators in real time;

[0081] Difference analysis module: calculates the dynamic difference characteristics of each actuator relative to the reference trajectory;

[0082] Adaptive weight allocator: Based on dynamic difference characteristics and health indicators, it generates fusion weights through fuzzy control and neural network;

[0083] Bias generation module: calculates bias instructions based on displacement difference, static pressure difference residual and weight;

[0084] Main instruction correction module: superimposes the offset instruction on the flight control main instruction to generate the final actuator instruction;

[0085] Fault isolation module: performs privilege reduction and output isolation operations on abnormal channels.

[0086] The dynamic difference characteristics output by the difference analysis module include static pressure difference residual, dynamic response difference, displacement difference, output force mutation and state instability index. The bias generation module performs the operation Δcmd i =K i ×(ΔX i +ΔP i ), where K i Dynamically adjusted by an adaptive weight allocator.

[0087] The sensor group (displacement encoder, pressure gauge) and signal conditioning circuit build a state acquisition module; the multi-core processor executes ΔX i ,ΔP i ,ΔT i Equal feature calculations are used to construct a difference analysis module; FPGA implements a fuzzy rule base or neural network reasoning to form an adaptive weight distributor; the multiplier and adder hardware execute Δcmd i =K i ×(ΔX i +ΔP i ) constitutes a bias generation module; the addition circuit superimposes u i =u nominal +Δ cmdi The main instruction correction module is formed; the comparator monitors the health threshold and triggers the weight clearing signal to form the fault separation module. The above six modules can interact through the high-speed bus during use. The difference analysis module outputs dynamic features to the weight distributor, which generates K i The error is then sent to the bias generation module, and ultimately output to the actuator driver by the main instruction correction module. The fault isolation module directly interrupts the data flow of the abnormal channel, enabling hardware acceleration (FPGA) to meet the real-time requirements of flight control (processing cycle ≤ 1ms). Its modular design is compatible with existing FBW systems, requiring only additional boards or software upgrades.

[0088] In summary, the advantages of the present invention are: through multi-dimensional state fusion and adaptive weight distribution, real-time suppression of internal force conflicts and smooth separation of fault channels are achieved, thereby achieving the effects of improving rudder control accuracy, reducing structural losses and ensuring fault tolerance.

[0089] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for controlling actuator force contention based on adaptive state fusion, characterized in that: include: Real-time collection of status data of multiple redundant actuators, including output displacement, output force or pressure difference, response delay, status noise and health indicators; Based on the reference trajectory and the state data of each actuator, the static pressure difference residual, dynamic response difference, displacement difference, output force mutation and state instability index are generated, and the dynamic difference characteristics are calculated; According to the dynamic difference characteristics and actuator health indicators, the fusion weight of each actuator is calculated through fuzzy control, neural network or rule logic dynamics; Based on the displacement difference, static pressure difference residual and weight, the bias compensation instruction is generated according to Δcmd i =K i ×(ΔX i +ΔP i ) generates independent compensation for each channel, where K i is the dynamic adjustment coefficient; Superimpose the offset command on the flight control main command, correct the main control command, and generate the final actuator command; When the dynamic residual of an actuator is abnormal or the health index is lower than the threshold, fault separation is performed, its weight is reduced and it is prohibited from participating in the control output.

2. The actuator force conflict control method based on adaptive state fusion according to claim 1, characterized in that: The dynamic difference features include: Static pressure difference residual: the steady-state deviation between the actual output force or pressure difference of each actuator and the target value; Dynamic response difference: The actuator response delay time difference estimated by step response; Displacement difference: The deviation between the actual displacement of the actuator and the reference trajectory, which includes the average displacement or target velocity trajectory; Output force mutation: the instantaneous rate of change of the actuator output force; State instability index: The statistical value of the actuator state noise calculated based on the signal fluctuation rate.

3. The actuator force conflict control method based on adaptive state fusion according to claim 2, characterized in that: The fusion weight is calculated in at least one of the following ways: Fuzzy control: output weights based on the fuzzy rule base of dynamic difference characteristics and health indicators; Rule logic: Dynamically assign weights through preset threshold logic or empirical rules; Neural network: Use training models to perform nonlinear mapping on multi-dimensional features to generate weights.

4. The actuator force conflict control method based on adaptive state fusion according to claim 3 is characterized in that: The generation formula of the offset compensation instruction is Δcmd i =K i ×(ΔX i +ΔP i ), where K i The coefficient of real-time adjustment is obtained through fuzzy control or neural network dynamic optimization based on dynamic difference characteristics, ΔX i is the current channel displacement difference, ΔP i is the static pressure difference residual.

5. The actuator force conflict control method based on adaptive state fusion according to claim 4 is characterized in that: The fault isolation process includes: When the dynamic residual continues to exceed the set threshold or the health indicator is lower than the safety lower limit; Reduce the weight of the actuator to zero and prohibit it from participating in the control output; Perform soft privilege reduction or complete isolation operations to ensure system fault tolerance.

6. The actuator force conflict control method based on adaptive state fusion according to claim 5, characterized in that: The health indicator is derived through built-in self-test or accumulated performance degradation data, where the built-in self-test is a real-time diagnosis of the electrical or mechanical performance of the actuator, and the accumulated performance degradation data is a health score based on historical status data, which includes response delay and noise growth.

7. The actuator force conflict control method based on adaptive state fusion according to claim 6, characterized in that: The reference trajectory is the average value of the displacement feedback of each actuator or the target speed trajectory issued by the flight control system.

8. A device for implementing the method according to any one of claims 1 to 7, characterized in that: include: State acquisition module: acquires the actuator's output displacement, output force or pressure difference, response delay, state noise and health indicators in real time; Difference analysis module: calculates the dynamic difference characteristics of each actuator relative to the reference trajectory; Adaptive weight allocator: Based on dynamic difference characteristics and health indicators, it generates fusion weights through fuzzy control and neural network; Bias generation module: calculates bias instructions based on displacement difference, static pressure difference residual and weight; Main instruction correction module: superimposes the offset instruction on the flight control main instruction to generate the final actuator instruction; Fault isolation module: performs privilege reduction and output isolation operations on abnormal channels.

9. The device for implementing the method according to any one of claims 1 to 7 according to claim 8, characterized in that: The dynamic difference features output by the difference analysis module include static pressure difference residual, dynamic response difference, displacement difference, output force mutation and state instability index.

10. The device for implementing the method according to any one of claims 1 to 7 according to claim 9, characterized in that: The bias generation module performs the operation Δcmd i =K i ×(ΔX i +ΔP i ), where K i Dynamically adjusted by an adaptive weight allocator.

Images