A method for constructing a composite controller for a collaborative robot joint system

Abstract

The application discloses a kind of construction methods of cooperative robot joint system composite controller, by initial value PI control module and extreme value optimization adaptive PI control module constitute input actual speed, speed error and dq axis current, output is the speed controller of control voltage;Position sensorless control module is constituted by sliding mode observer and arctangent function, by nonlinear tracking differentiator, nonlinear state error feedback controller, neural network compensator, two adjustable gains and three-order extended state observer constitute position controller;Actual speed is output by rotor position angle according to first-order differential tracker, and composite controller is jointly constituted by position controller, speed controller and first-order differential tracker;The application will conventional three closed-loop control be simplified into position-current double closed-loop, while being combined with sensorless control method, so that the requirement of control system to external hardware and motor parameters is reduced, while having very strong static, dynamic characteristics and robustness.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of intelligent robot motion control, and more particularly relates to a controller construction method for a robot flexible joint system. BACKGROUND

[0002] With the continuous development of robot technology, collaborative robots, as an advanced system with flexibility and intelligence, have gradually become a research hotspot. The joint system of a robot, as a core component connecting the mechanical structure and the control system, plays a key role in achieving complex motion tasks of the robot. The design and optimization of the joint system controller have a great impact on the performance and application of collaborative robots.

[0003] The current conventional control method for the joint system is generally a "position-velocity-current" three-closed-loop control system based on a PI module. The use of multiple PI controllers not only makes the control system have many parameters and be difficult to tune, but also greatly affects the dynamic performance, stability, and robustness of the system. In view of these problems, the document with the Chinese patent publication number CN117532606A provides a joint servo system control method based on an improved active disturbance rejection controller. The method designs an extended state observer based on the linear active disturbance rejection theory, and designs a new function to observe the uncertain terms of the system. However, the linear active disturbance rejection controller does not compensate for the inherent defects of the observer, resulting in errors in the control. In the document "Gao Xusheng. Design and implementation of permanent magnet synchronous motor position control system based on active disturbance rejection [D]. Yangzhou University, 2024.", a "position-current" double-closed-loop control system is constructed based on an active disturbance rejection controller. The output speed position of the tracking differentiator is controlled through a PI controller. However, this method does not consider the shortcomings of the linear active disturbance rejection controller, and the PI controller is not designed in close combination with the control system of the motor. Therefore, the above two schemes cannot meet the high-precision, high-robustness, and high-dynamic performance requirements of collaborative robots, and the controller needs to be optimized in combination with the specific control system. SUMMARY

[0004] The present application aims to solve the above-mentioned problems of the current conventional control method for the joint system and provide a construction method for a composite controller of a collaborative robot joint system. The composite controller constructed by the method is based on nonlinear active disturbance rejection control and adaptive PI algorithm, and can meet the high-precision, high-robustness, and high-dynamic performance requirements of collaborative robots.

[0005] To achieve the above-mentioned purpose, the construction method for a composite controller of a collaborative robot joint system adopted by the present application includes the following steps:

[0006] Step 1): a speed controller with input of actual speed speed error e ω and dq axis current i d , q and output of control voltage u d , q ;

[0007] The output and input relationship of the initial value PI control module is:

[0008]

[0009] u PI0 represents a voltage matrix, and respectively represent two sets of initial proportional-integral coefficients, d-axis error current L d is the inductance of the d-axis, and R s is the resistance;

[0010] The output and input relationship of the extreme value optimization adaptive PI control module is:

[0011]

[0012] Two sets of PI parameters and are obtained based on the sliding mode control law and the gradient descent method, and are minimized under the stable condition;

[0013] The control voltage

[0014] Step 2): a position sensorless control module composed of a sliding mode observer and an arctangent function is constructed, with input of αβ-axis voltage v α , β and current i α , β , and output of rotor position angle

[0015] Step 3): a composite controlled object is composed of joint motors, a position sensorless control module, a 2s / 2r transformation module, an SVPWM module, an inverter, an electric sensor, a 3s / 2s transformation module, and a 2r / 2s transformation module, with input of control voltage u d , q , and output of rotor position angle and current i d , q ;

[0016] Step 4): the position controller is composed of a nonlinear tracking differentiator, a nonlinear state error feedback controller, a neural network compensator, two adjustable gains and a third-order extended state observer, the position controller outputs a reference position angle * and a rotor position angle Step 5): the first-order differential tracker outputs an actual rotor speed

[0017] Step 5): the first-order differential tracker outputs an actual rotor speed Step 5): the first-order differential tracker outputs an actual rotor speed Step 5): the first-order differential tracker outputs an actual rotor speed Step 5): the first-order differential tracker outputs an actual rotor speed Step 5): the first-order differential tracker outputs an actual rotor speed ω The cooperative robot joint system composite controller is composed of the position controller, the speed controller and the first-order differential tracker.

[0018] Further, the extreme value optimization adaptive PI control module adopts an extreme value optimization adaptive PI algorithm based on a sliding mode control law, and uses two sliding mode surfaces K1 and K2 are two positive sliding mode gain coefficients, and the two groups of PI parameters and The expression is:

[0019] σ is the learning rate of the gradient descent method, the superscript indicates the number of groups of PI parameters, and the subscript indicates the type corresponding to the learning rate.

[0020] Further, the output of the sliding mode observer is an observed extended back electromotive force and k is a sliding mode gain coefficient, and is a predicted current; the extended back electromotive force and The rotor position is calculated through an arctangent function

[0021] The present application has the following advantages:

[0022] 1. The position controller in the application controls the position through an optimized nonlinear active disturbance rejection controller, observes two position signals and a speed signal according to total disturbance of the system through a third-order extended state observer, and compensates the speed signal through a neural network, thereby reducing the accuracy requirement for system parameters, and making the position controller still have higher control accuracy, dynamic and static performance and anti-interference ability under external interference.

[0023] 2. The speed controller in this invention controls the rotational speed through an extreme value optimization composite PI controller. Through redesign, the PI initial value module can directly output the initial value of the control voltage based on the observed rotational speed, rotational speed error, and current signal. An extreme value optimization algorithm based on sliding mode control and gradient descent method is proposed to correct the parameters based on the PI initial value. The control voltage is output directly based on the reference rotational speed, avoiding redundant adjustment of the current loop PI parameters. It can dynamically adapt to system changes, improve stability and robustness, and provide stable and reliable operation under changes in motor parameters.

[0024] 3. This invention simplifies the conventional three-loop control into a "position-current" dual-loop control while combining it with a sensorless control method. This reduces the requirements of the control system on external hardware and motor parameters, while providing strong static and dynamic characteristics and robustness to external disturbances, making it promising for wide application in engineering practice. The proposed construction method is of great significance to the research and application of collaborative robot joint system control, both academically and practically. Attached Figure Description

[0025] Figure 1 This is a block diagram of a composite controller for a collaborative robot joint system constructed using the method of this invention.

[0026] Figure 2 yes Figure 1 A block diagram of the composition of a composite controlled object;

[0027] Figure 3 yes Figure 2 Block diagram of the sensorless control module;

[0028] Figure 4 yes Figure 1 Block diagram of the position controller;

[0029] Figure 5 yes Figure 1 Block diagram of the medium speed controller. Detailed Implementation

[0030] The present invention will now be described in further detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0031] See Figure 1 The method of this invention constructs a composite controller for a collaborative robot joint system, which consists of a position controller, a velocity controller, and a first-order differential tracker. This controller is connected to a composite controlled object, including joint motors but without a position sensor control module, to achieve control of the composite controlled object, i.e., the collaborative robot joint system. The input to the composite controlled object is a control voltage u.d , u q , output rotor position angle and current i d , i q ; rotor position angle is input into a position controller and a first-order derivative tracker respectively, and dq-axis current i d , i q is input into a speed controller. The position controller outputs a reference rotational speed * according to reference position angle and rotor position angle The first-order derivative tracker outputs an actual rotational speed according to rotor position angle Reference rotational speed is subtracted from actual rotational speed to obtain a rotational speed error Actual rotational speed and rotational speed error e ω are both input into the speed controller. The speed controller obtains dq-axis control voltage u d , u q according to input actual rotational speed rotational speed error e ω and dq-axis current i d , i q to control the compound controlled object. Specifically as follows:

[0032] Referring to Figure 3 , a position sensorless control module is constructed. The position sensorless control module is composed of a sliding mode observer and an arctangent function. The input of the position sensorless control module is αβ-axis voltage v α , v β and current i α , i β , and the output is rotor position angle The αβ-axis voltage v α , v β and current i α , i β output by the joint motor are input into the position sensorless control module. The joint motor outputs voltage signal v abc and current signal i abc to a 3s / 2r transformation module, and the 3s / 2r transformation module outputs αβ-axis voltage v α , v β and current i α , i β to the sliding mode observer, and the extended back electromotive force is observed. The output of the sliding mode observer is observed extended back electromotive force and The relationship between the output and input of the sliding mode observer can be expressed as follows:

[0033]

[0034] where k is the sliding mode gain coefficient, and is the predicted current; the predicted current and The differential signal of S

[0035]

[0036] where μ α and μ β is the sliding mode control law in the stationary reference frame, S α and S β is the sliding surface in the stationary reference frame; A is a coefficient matrix related to the motor parameters, which can be expressed as follows:

[0037]

[0038] where L d is the d-axis inductance, L q is the q-axis inductance, R s is the resistance, and ω e is the electrical angular velocity, which can be calculated by the power supply frequency f;

[0039] ω e = 2πf (4)

[0040] The predicted current and can be expressed as follows:

[0041]

[0042] The extended back electromotive force and The observed joint motor rotor position angle

[0043]

[0044] Combining Figure 2As shown, a composite controlled object is constructed from a joint motor, a sensorless control module, a 2S / 2R conversion module, an SVPWM module, an inverter, an electrical sensor, a 3S / 2S conversion module, and a 2R / 2S conversion module. Precise control of the joint motor is achieved based on the input voltage signal. The 2S / 2R conversion module, SVPWM module, inverter, electrical sensor, and 3S / 2S conversion module are connected sequentially. The output of the 3S / 2S conversion module is connected to both the 2R / 2S conversion module and the sensorless control module. The electrical sensor is connected to the joint motor. The input of the 2S / 2R conversion module is the dq-axis voltage u. d u q After 2s / 2r transformation, it is converted into an αβ axis voltage u. α u β It is then output to the SVPWM module, and its transformation relationship is as follows:

[0045]

[0046] Voltage u α u β Six switching signals are output through the SVPWM module. 1～6 Six-channel switch signal s 1～6 The input is to the inverter, which receives a DC voltage V. dc Powered by a motor that drives the joint.

[0047] The electrical sensor acquires the voltage v output by the inverter using digital signals. a v b v c Signal and current i a i b i c Signal, i.e., v abc Signal and i abc The signal is input to the 3s / 2s conversion module, and after 3s / 2s conversion, the voltage v of the αβ axis is output. α v β and current i α i β The relationship between the output and the input is expressed as follows:

[0048]

[0049] Current i along the αβ axis α i β The input current i in the 2r / 2s conversion module is the output current i after the 2r / 2s conversion. d i q ;

[0050]

[0051] The current signal i d , i q output by the 2s / 2r conversion module is fed back to the speed controller in Figure 1 .

[0052] The voltage v α , v β and the current i α , i β of the αβ axis are input to the position sensorless control module in Figure 3 , and the rotor position angle of the joint motor is output after position sensorless control, and the rotor position angle is fed back to the 2s / 2r conversion module and the 2r / 2s conversion module.

[0053] The observed rotor position is output by the position sensorless control module to the position controller, the first-order differential tracker in Figure 1 , and each coordinate conversion module of the composite controlled object, so as to realize position feedback.

[0054] Referring to Figure 1 , the relationship between the input rotor position angle of the first-order differential tracker and the output actual speed can be expressed as follows:

[0055]

[0056] Referring to Figure 4 , the position controller is constructed, and the position controller is composed of a nonlinear tracking differentiator, a nonlinear state error feedback controller, a neural network compensator, two adjustable gains, and a third-order extended state observer. The input of the nonlinear tracking differentiator is the given reference position angle θ * , and the output is two position differential signals and wherein is used to track the reference position angle θ * , as the approximate differential of θ * , can be expressed as The relationship between the differential signal of the output of the nonlinear tracking differentiator and the input is as follows:

[0057]

[0058] wherein h is an adjustable filtering factor, generally taken as 0.1-10; r is an adjustable tracking speed factor, generally taken as 0.1-1; the values of the adjustable factors need to be selected according to the control effect of the system, and fhan is a fastest control synthesis function.

[0059] The position signal and The signal of the above equation can be expressed as follows:

[0060]

[0061] A third-order extended state observer is constructed. The third-order extended state observer estimates the disturbances inside and outside the system in real time, and compensates for these disturbances through feedback control, thereby effectively eliminating the influence caused by the disturbances in the system, so that the system has strong anti-interference ability. The input of the third-order extended state observer is the rotational speed ω2 and the observed rotor position angle of the output of the composite controlled object The output is three state variables z1, z2 and z3, wherein z1, z2 are used to track the position signal of the output of the composite controlled object, and z3 is used to track the total unknown disturbance of the rotational speed of the system; the relationship between the differential signal of the output of the third-order extended state observer and the input is expressed as follows:

[0062]

[0063] Wherein β1, β2 and β3 are three adjustable gain coefficients of the third-order extended state observer; α1, α2 and α3 are adjustable parameters of three fal functions, generally taking 0-1; δ1, δ2 and δ3 are three adjustable parameters of the fal function, generally taking 5T s ≤ δ ≤ 10T s , wherein T s is the sampling time of the digital controller.

[0064] The nonlinear fal function is expressed as follows:

[0065]

[0066] Wherein x is the observed state variable; α and δ are adjustable parameters of the fal function.

[0067] Two position differential signals and output by the nonlinear tracking differentiator are respectively subtracted from two observed angle z1, z2 signals output by the third-order extended state observer, to obtain two error signals and

[0068] A nonlinear state error feedback controller is constructed. Through two error signals and The disturbance is inhibited and reduced, so that the system can more accurately respond to a given signal while effectively resisting the influence of external disturbance. The input of the nonlinear state error feedback controller is two error signals e1 and e2 obtained according to the nonlinear tracking differentiator and the third-order extended state observer; the output of the nonlinear state error feedback controller is a speed ω0 signal; the relationship between the output and the input of the nonlinear state error feedback controller is expressed as follows:

[0069] ω0=λ1fal(e1,α4,δ4)+λ2fal(e2,α5,δ5) (15)

[0070] Wherein, λ1 and λ2 are two error gain coefficients, α4 and α5, δ4 and δ5 are adjustable parameters of the fal function, and the value range is the same as α 1～3 and δ 1～3 ;

[0071] The neural network compensator is constructed. The total disturbance is compensated according to the variables output by the third-order extended state observer, so that the accuracy and robustness of the control system are improved; the input of the neural network compensator is two angle signals z1 and z2 output by the third-order extended state observer; the output of the neural network compensator is two compensation signals c1 and c2; taking two layers as an example, the relationship between the output and the input of the neural network compensator based on the multilayer perceptron is expressed as follows:

[0072]

[0073] Wherein, h1 and h2 are the outputs of the hidden layer; are weight coefficients of neurons, the superscript represents the layer number, and in the present application, 1 and 2 are taken; the subscript represents the serial number of the neuron in the layer and the serial number of the neuron connected to the next layer, and in the present application, 11, 12, 21 and 22 are taken; τ is the bias of the neuron, the superscript represents the layer number, and in the present application, 1 and 2 are taken; the subscript represents the serial number of the neuron, and in the present application, 1 and 2 are taken; f(·) and g(·) are activation functions of the hidden layer and the output layer respectively, which can be selected as sigmoid, ReLU, tanh function, etc.

[0074] The angle compensation signal c2 output by the neural network compensator is superimposed with the speed signal ω0 output by the nonlinear state error feedback controller through an adjustable gain to obtain a speed signal ω2; ω2 can be expressed as follows:

[0075] ω2=bω0+c2 (17)

[0076] Wherein b is a gain coefficient.

[0077] The compensation signal c1 output by the neural network compensator is superimposed with the observed speed signal z3 output by the third-order extended state observer and the speed signal ω1 output by the variable gain 1; ω1 can be expressed as follows:

[0078]

[0079] Where b is a gain coefficient.

[0080] The speed signal ω1 output by the adjustable gain is subtracted from ω0 output by the nonlinear state error feedback controller to obtain the reference speed ωref output by the position controller which can be expressed as follows:

[0081]

[0082] The nonlinear tracking differentiator, the nonlinear state error feedback controller, the third-order extended state observer, and the neural network compensator are sequentially connected to form a position controller, and according to a given reference position angle θref * and the observed rotor position output by the composite controlled object, the reference speed ωref is output.

[0083] As shown in Figure 5 , the speed controller is composed of an initial value PI control module and an extreme value optimization adaptive PI control module. The speed controller based on the extreme value optimization adaptive composite PI controller is constructed to replace the speed loop PI controller and the current loop PI controller in the conventional control scheme, which simplifies the structure of the control system and avoids the increase in cost caused by redundant control links. The input of the speed controller is the speed error the current signal i d output by the composite controlled object, q the actual speed ω1 output by the first-order differential tracker, and the output of the speed controller is the voltage signal u d . q .

[0084] The relationship between the output of the initial value PI control module and the input can be expressed as follows:

[0085]

[0086] Where u PIO represents the output voltage matrix of the initial value PI control module. and respectively represent two groups of initial proportional-integral coefficients in the initial value PI control module, and the values can be obtained according to calculation and adjusted according to actual effect, the approximate range of the proportional coefficient is 0.1-100, and the range of the integral coefficient is 0.01-10; i e is the d-axis error current, which can be expressed as follows:

[0087]

[0088] In order to adaptively adjust the initial PI coefficient, an extreme value optimization adaptive PI algorithm based on a sliding mode control law is proposed; two sliding mode surfaces used in the application are selected as follows:

[0089]

[0090] Wherein, K1 and K2 are two positive sliding mode gain coefficients.

[0091] According to the gradient descent method, the minimum stable condition is minimized to obtain the first-order differential expression of the adaptive PI coefficient under stability as follows:

[0092]

[0093] Wherein, sigma is the learning rate of the gradient descent method, the superscript indicates the number of PI parameters, and the subscript indicates the type of learning rate; And Two groups of PI parameters in the extreme value optimization adaptive PI control module are expressed as follows:

[0094]

[0095] The relationship between the output and the input of the extreme value optimization adaptive PI control module is expressed as follows:

[0096]

[0097] Further, the relationship between the output and the input of the speed controller can be expressed as follows:

[0098]

[0099] Then:

[0100]

[0101] Wherein, is in the form of a matrix of dynamic error, which can be expressed as follows:

[0102]

[0103] Wherein, mu1 and mu2 are positive gain coefficients.

[0104] As shown in Figure 1 The position controller, the speed controller and the first-order differential tracker are connected in sequence to form a composite controller, and the motion of the articulated robot joint system is completed according to the input voltage signal.

[0105] The above merely describes the preferred embodiments of the present application and is not used to limit the present application. Any modification, equivalent replacement and improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A configuration method of a cooperative robot joint system composite controller characterized by The method comprises the following steps: Step 1): a speed controller with input of actual rotating speed rotating speed error e ω and dq-axis current i d , q and output of control voltage u d , q composed of initial value PI control module and extreme value seeking adaptive PI control module The output-input relationship of the initial value PI control module is: u PI0 represents a voltage matrix, and represent two sets of initial proportional-integral coefficients, respectively, d-axis error current L d is the inductance of the d-axis, R s is the resistance; The output-input relationship of the extreme value optimization adaptive PI control module is: Two sets of PI parameters and are obtained based on the sliding mode control law, according to the gradient descent method, and the stability condition The control voltage Step 2): construction of the position sensorless control module consisting of a sliding mode observer and an arctangent function, whose inputs are the voltages v α β and currents i α β in the a and b axes, and whose output is the rotor position angle ​​ Step 3): the composite controlled object is composed of joint motor, position sensorless control module, 2s / 2r conversion module, SVPWM module, inverter, electric sensor, 3s / 2s conversion module and 2r / 2s conversion module, the input of the composite controlled object is control voltage u d , q , the output is rotor position angle and current i d , q ; Step 4): The position controller is composed of a nonlinear tracking differentiator, a nonlinear state error feedback controller, a neural network compensator, two adjustable gains, and a third-order extended state observer. The position controller generates the reference angular velocity * and the reference angular velocity ​ Step 5) The first-order derivative tracker according to the rotor position angle Output actual rotation speed Reference rotation speed Subtract the actual rotation speed The rotation speed error e is obtained ω The position controller, the speed controller, and the first-order derivative tracker jointly constitute a cooperative robot joint system composite controller.

2. The method for constructing a composite controller for a collaborative robot joint system according to claim 1, characterized in that: The extreme value optimization adaptive PI control module adopts an extreme value optimization adaptive PI algorithm based on a sliding mode control law, uses two sliding mode surfaces K1 and K2 are two positive sliding mode gain coefficients, and the two sets of PI parameters and The expression is: σ is the learning rate of the gradient descent method, the superscript indicates the number of PI parameters, and the subscript indicates the type to which the learning rate corresponds.

3. The method of claim 1, wherein the method further comprises: providing a plurality of sensors configured to provide sensor data to the controller; and providing a plurality of actuators configured to receive actuation commands from the controller. The output of the sliding mode observer is an estimated extended back emf and k is a sliding mode gain coefficient, and is a predicted current; an extended back emf and The rotor position is calculated by an arctangent function 4. The method of claim 3, wherein the method further comprises: providing a plurality of sensors configured to provide sensor data to the controller; and providing a plurality of actuators configured to receive actuation commands from the controller. The predicted current And The differential equation is: μ α And μ β Is the sliding mode control law under the stationary reference frame, S α And S β Is the sliding surface under the stationary reference frame; A is the coefficient matrix related to motor parameters A is L d Is the inductance of the d-axis, L q Is the q-axis inductance, R s Is the resistance, ω e Is the electrical angular velocity.

5. The method of constructing a cooperative robotic joint system composite controller of claim 1, wherein: The output-input relationship of the differential signal of the output of the nonlinear tracking differentiator is: h is an adjustable filter factor, r is an adjustable tracking speed factor, and fhan is a fastest control comprehensive function.

6. The method for constructing a composite controller for a collaborative robot joint system according to claim 5, characterized in that: The output-input relationship of the differential signal of the output of the third-order extended state observer is: z1, z2 and z3 are three state variables, β1, β2 and β3 are three adjustable gain coefficients; α1, α2 and α3 are three adjustable parameters of the fal function, δ1, δ2 and δ3 are three adjustable parameters of the fal function, take 5T s ≤ δ ≤ 10T s , T s is the sampling time; The non-linear fal function is: x is the observed state quantity; and a and d are adjustable parameters of the fal function.

7. The method of constructing a cooperative robotic joint system composite controller of claim 6, wherein: The output-input relationship of the rotational speed ω0 signal of the output of the linear state error feedback controller is: ω0 = λ1fal(e1, α4, δ4) + λ2fal(e2, α5, δ5), λ1 and λ2 are two error gain coefficients, and α4 and α5, δ4 and δ5 are adjustable parameters of the fal function, 8. The method of claim 7, wherein the method further comprises: providing a plurality of sensors configured to provide sensor data to the controller; and providing a plurality of actuators configured to receive actuation commands from the controller. The output-input relationship of the two angle compensation signals c1 and c2 of the neural network compensator is: h1 and h2 are the outputs of the hidden layers; are the weight coefficients of the neurons, the superscripts indicate the layer number, taking 1 and 2; the subscripts indicate the sequence number of the neuron in the current layer and the sequence number of the neuron connected to the next layer, taking 11, 12, 21 and 22; τ are the bias quantities of the neurons, the superscripts indicate the layer number, taking 1 and 2; the subscripts indicate the sequence number of the neuron, taking 1 and 2; f(·) and g(·) are the activation functions of the hidden layer and the output layer, respectively.

9. The method of claim 8, wherein the method further comprises: providing a plurality of sensors configured to provide sensor data to the controller; and providing a plurality of actuators configured to receive actuation commands from the controller. The angle compensation signal c2 is superimposed with the rotation speed ω0 through a first adjustable gain output to obtain a rotation speed ω2 = bω0 + c2, b being a gain coefficient; the angle compensation signal c1 is superimposed with the rotation speed state variable z3 through a second variable gain output to obtain a rotation speed ω1 = aω0 + c1, a being a gain coefficient reference rotation speed 10. The method of claim 1, wherein the method further comprises: providing a plurality of sensors configured to provide sensor data to the controller; and providing a plurality of actuators configured to receive actuation commands from the controller. 2s / 2r conversion module, SVPWM module, inverter, electric sensor and 3s / 2s conversion module are connected in sequence, the output of 3s / 2s conversion module is connected with 2r / 2s conversion module and position sensorless control module respectively, the electric sensor is connected with joint motor, the input of 2s / 2r conversion module is dq axis voltage u d , u q , which is converted into αβ axis voltage u α , u β by 2s / 2r conversion and output to SVPWM module, SVPWM module outputs six-way switching signal s 1～6 , which is input into inverter, the inverter is driven by direct current voltage to drive joint motor; the electric sensor collects voltage v a , v b , v c , current i a , i b , i c signal output by inverter and inputs into 3s / 2s conversion module, the voltage v α , v β and current i α , i β , i α , i β output by 3s / 2s conversion are input into 2r / 2s conversion module, the current i d , i q output by 2r / 2s conversion is output; the current i d , i q output by 2s / 2r conversion module is input into speed controller for feedback, the voltage v α , v β and current i α , i β are input into position sensorless control module, the rotor position angle output by position sensorless control is output, and the rotor position angle is fed back to 2s / 2r conversion module and 2r / 2s conversion module.

Images