A dual-arm robot adaptive control method based on mirror principle, electronic equipment and storage medium

Abstract

The application belongs to the technical field of dual-arm robot control, and particularly discloses a dual-arm robot adaptive control method based on a mirror principle, an electronic device and a storage medium, which comprises the following steps: S1, obtaining parameters of healthy-side and affected-side mechanical arms; S2, constructing healthy-side and affected-side mechanical arm state quantities at a current time; S3, introducing a collision coefficient C ; S4, constructing healthy-side and affected-side mechanical arm dynamics models to obtain an auxiliary force of a prediction model; S5, cyclically executing S1-S4 to introduce an optimization function R ( X ( t )); S6, cyclically executing S1-S5 to introduce a weight coefficient adjustment factor β ; S7, cyclically executing S1-S6 to dynamically adjust adaptive weight coefficients, prediction model weight coefficients and the optimization function; S8, real-time adjusting a minimum auxiliary force size and direction of the affected-side mechanical arm; S9, collecting and processing data of the healthy-side and affected-side mechanical arms of the patient, and generating a table; and S10, evaluating overall recovery of the patient in the rehabilitation process.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of dual-arm robots, and particularly relates to a dual-arm robot adaptive control method based on a mirror principle, an electronic device and a storage medium. BACKGROUND

[0002] Traditional rehabilitation training usually relies on the auxiliary operation of medical staff, and has problems of low efficiency, difficulty in personalization, inability to adjust the auxiliary force in real time, etc. Adopting minimum auxiliary force to assist patients in rehabilitation has many advantages and effects. Adopting minimum auxiliary force control can reduce excessive intervention in the rehabilitation process of patients, help patients establish more autonomous motor ability, promote the recovery of muscle strength and joint range of motion, and thus improve the rehabilitation effect and treatment efficiency. The existing dual-arm robot adopts a constant auxiliary force mode when the healthy-side mechanical arm assists the affected-side mechanical arm, and cannot dynamically adjust and ensure safety according to individual differences and specific movement conditions of patients, so that the rehabilitation effect is difficult to reach the best state.

[0003] In addition, the existing dual-arm robot adaptive control strategy based on the mirror principle has obvious defects such as lack of individual adaptability, untimely auxiliary force adjustment, and insufficient collision risk prevention. SUMMARY

[0004] In order to solve the above technical problems, the application provides a dual-arm robot adaptive control method based on a mirror principle, an electronic device and a storage medium.

[0005] Based on the above purpose, the application realizes the following technical scheme:

[0006] The first aspect of the application provides a dual-arm robot adaptive control method based on a mirror principle, comprising the following steps:

[0007] S1, start a mirror training task scene such as shooting, the patient is located at an initial position and holds the end handle of the dual-arm robot; according to the shooting task, based on the mirror principle, the healthy-side mechanical arm actively drives the affected-side mechanical arm to perform rehabilitation training movement; the relevant data of the healthy-side and affected-side mechanical arms are acquired in real time through joint angle sensors and force sensors, and the parameters of the real-time position x, velocity v, acceleration a and interaction force F of the end TCP of the healthy-side and affected-side mechanical arms are obtained after processing.

[0008] S2, according to the parameters of the healthy side and the affected side of the robot arm position x, velocity v, acceleration a, interaction force F obtained in S1, the state quantity Xhealthy(t) of the healthy side robot arm and the state quantity Xaffected(t) of the affected side robot arm at the current moment are constructed, the difference between the healthy side robot arm and the affected side robot arm at the current moment is obtained, and the state quantity Xhealthy(t) of the healthy side robot arm and the state quantity Xaffected(t) of the affected side robot arm respectively include the state quantity of the position, velocity, acceleration, interaction force and other parameters of the healthy side and the affected side robot arm.

[0009] S3, according to the parameters of the healthy side and the affected side of the robot arm obtained in S1, input into the adaptive control network model of the dual-arm robot, real-time monitor the change of the position x, velocity v, acceleration a, interaction force F and other parameters of the current healthy side and the affected side of the robot arm, introduce the collision coefficient C, which is used to predict the collision risk of the healthy side and the affected side of the robot arm, by discretizing the connecting rod of each robot arm into a certain number of points, and calculating the Euclidean distance d between each point of the other side robot arm and the current robot arm point, the minimum value d min is screened out, the safety distance Z between the healthy side and the affected side of the robot arm is defined, and whether the minimum value d min of the Euclidean distance between the healthy side and the affected side of the robot arm is less than the safety distance Z is judged in real time, if it is less than the safety distance, there is a collision risk, then C=-1, otherwise C=1.

[0010] S4, according to the state quantity Xhealthy(t) of the healthy side robot arm and the state quantity Xaffected(t) of the affected side robot arm at the current moment, input into the adaptive control network model of the dual-arm robot, and use these information to construct the dynamics model of the healthy side and the affected side of the robot arm, predict the predicted quantity Nhealthy(t+1) of the healthy side robot arm at the next moment and the predicted quantity Naffected(t+1) of the affected side robot arm at the next moment, obtain the auxiliary force of the prediction model, and the predicted quantity Nhealthy(t+1) of the healthy side robot arm at the next moment and the predicted quantity Naffected(t+1) of the affected side robot arm at the next moment respectively include the state quantity of the predicted position, velocity, acceleration, interaction force and other parameters of the healthy side and the affected side of the robot arm.

[0011] S5, in the adaptive control strategy of the dual-arm robot, S1-S4 are executed circularly, the state quantity Xhealthy(t) of the healthy side robot arm, the state quantity Xaffected(t) of the affected side robot arm, the predicted quantity Nhealthy(t+1) of the healthy side robot arm at the next moment and the predicted quantity Naffected(t+1) of the affected side robot arm at the next moment are continuously monitored and recorded, and the optimization function R(X(t)) is introduced, which is used to optimize the motion effect of the dual-arm robot and the individualization degree of the rehabilitation training, so as to optimize the control strategy and improve the rehabilitation effect.

[0012] S6, circulating S1-S5, based on the obtained multiple sets of data, a weight coefficient adjustment factor β is introduced as a fixed scaling coefficient of each weight term, used to control the amplification or reduction of the overall weight, the larger the value of β, the higher the sensitivity and response speed of the system, the smaller the value of β, the lower the sensitivity but the higher the stability of the system.

[0013] S7, circulating S1-S6, by real-time acquisition of the position x, velocity v, acceleration a, interaction force F parameters of the healthy side and the affected side mechanical arm end and the auxiliary force of the prediction model, according to the difference change of the healthy side and the affected side mechanical arm data, dynamically adjusting the adaptive weight coefficient, the prediction model weight coefficient and the optimization function.

[0014] S8, according to the dynamic change of each weight coefficient and the optimization function, real-time adjustment of the minimum auxiliary force size and direction of the affected side mechanical arm, used to ensure that the patient completes the rehabilitation training task with the minimum auxiliary force in the rehabilitation training process.

[0015] S9, after completing the mirror rehabilitation training task each time, the data of the patient's healthy side and affected side mechanical arm are collected and processed, and a table is generated, which is used to record the parameter difference between the healthy side mechanical arm and the affected side mechanical arm with time.

[0016] S10, based on the table generated by S9, comparing and analyzing the size and change trend of the difference between the healthy side and the affected side mechanical arm, evaluating the overall recovery of the patient in the rehabilitation process, and determining the effect after the first rehabilitation.

[0017] According to the above-mentioned adaptive control method of double-arm robot based on mirror principle, preferably, in step S2,

[0018] The Xhealthy(t) includes the position x d , velocity v d , acceleration a d and interaction force F d parameters of the healthy side mechanical arm at t time; the healthy side mechanical arm state quantity Xhealthy(t) = (x d , v d , a d , F d ).

[0019] The Xhealthy(t) includes the position x m , velocity v m , acceleration a m and interaction force F m parameters of the healthy side mechanical arm at t time; the healthy side mechanical arm state quantity Xhealthy(t) = (x m , v m , a m , F m ).

[0020] The difference ΔX between the healthy side and the affected side mechanical arm is:

[0021] ΔX = Xhealthy(t) - Xaffected(t);

[0022] ΔX = (x d -x m ,v d -v m ,a d -a m ,F d -F m )·

[0023] According to the above-mentioned adaptive control method of the dual-arm robot based on the mirror principle, preferably, in step S3, in order to effectively prevent the healthy side and the affected side mechanical arms from colliding during the training process, the step of introducing the collision coefficient C is: first discretize the connecting rod of each mechanical arm into a certain number of points, and calculate the Euclidean distance d between each point of the other side mechanical arm and the current mechanical arm point, and screen out the minimum value d min , define the safety distance Z between the healthy side and the affected side mechanical arms, and real-time monitor and judge whether the minimum value d min of the Euclidean distance between the healthy side and the affected side mechanical arms is less than the safety distance Z, if d min ≤ Z, it is a collision state, then the collision coefficient C = -1; if d min > Z, it is a non-collision state, then the collision coefficient C = 1, so as to change the direction of the mechanical arm auxiliary force, thereby effectively avoiding the collision between the healthy side and the affected side mechanical arms.

[0024] The Euclidean distance formula and the collision threshold setting formula are:

[0025]

[0026] If d min ≤ Z, set it as the collision threshold, determine the collision state, the collision coefficient C = -1, otherwise C = 1.

[0027] According to the above-mentioned adaptive control method of the dual-arm robot based on the mirror principle, preferably, in step S4,

[0028] The dynamics model of the healthy side mechanical arm is:

[0029]

[0030] Wherein, M 健 represents the mass matrix of the healthy side mechanical arm; represents the end acceleration of the healthy side mechanical arm; C 健 represents the damping matrix of the healthy side mechanical arm; represents the end velocity of the healthy side mechanical arm; K健 represents the stiffness property of the healthy-side manipulator; x 健 represents the end position of the healthy-side manipulator; F ext represents the force acting on the end of the healthy-side manipulator.

[0031] The dynamics model of the affected-side manipulator is:

[0032]

[0033] wherein M 患 represents the mass matrix of the affected-side manipulator; represents the end acceleration of the affected-side manipulator; C 患 represents the damping matrix of the affected-side manipulator; represents the end velocity of the affected-side manipulator; K 患 represents the stiffness matrix of the affected-side manipulator; x 患 represents the end position of the affected-side manipulator; F assistant represents the assisting force exerted by the robot on the affected-side manipulator; F interaction represents the interaction force between the affected-side manipulator and the external environment.

[0034] The state prediction of the next-time healthy-side manipulator prediction Nhealthy(t+1) and the next-time affected-side manipulator prediction Naffect(t+1) is:

[0035]

[0036] wherein X 健 (t) represents the state of the healthy-side manipulator at the current time; X 患 (t) represents the state of the affected-side manipulator at the current time; f and g represent functions for predicting the states of the healthy-side and affected-side manipulators, respectively.

[0037] The assisting force τ MPC is obtained by:

[0038]

[0039] wherein N 健 (t+1)” represents the acceleration prediction of the healthy-side manipulator at the next time; N 患 (t+1)” represents the acceleration prediction of the affected-side manipulator at the next time; N 健 (t+1)' represents the velocity prediction of the healthy-side manipulator at the next time; N 患 (t+1)') represents the velocity prediction of the affected-side manipulator at the next time; N 健 (t+1) represents the displacement prediction of the healthy-side manipulator at the next time; N 患 (t+1) represents the displacement prediction of the affected-side manipulator at the next time; M患 represents the mass matrix of the affected side robotic arm; C 患 represents the damping matrix of the affected side robotic arm; K 患 represents the stiffness matrix of the affected side robotic arm.

[0040] According to the adaptive control method of the dual-arm robot based on the mirror principle, preferably, in step S5, the state quantity X(t), the prediction quantity N(t) and the optimization function R(X(t)) of the patient's healthy side and affected side robotic arms are continuously monitored and recorded during the rehabilitation training task process, and the optimization function is:

[0041] R(X(t))=w1·e1+w2·e2;

[0042] wherein w1 and w2 represent weight coefficients, and w1+w2=1; e1 represents a rehabilitation effect index, e1>1, e1 includes personalized setting of joint range of motion size, muscle strength, and the better the rehabilitation effect, the higher the value of e1; e2 represents a safety index, e2>1, e2 is the number of times of collision risk warning between the healthy side and the affected side robotic arms, and the higher the value of e2, the better the rehabilitation training process.

[0043] The optimization function R(X(t)) is used to reduce unnecessary auxiliary force and play a damping role in sharp motion, further optimize the control strategy, and ensure that the patient's affected side robotic arm always assists the patient in rehabilitation with the smallest auxiliary force, while the value of R(X(t)) will increase sharply when encountering sudden and rapid mirror training motion;

[0044]

[0045] At this time, the auxiliary force F assistant is reversed, thereby playing a damping adjustment and slowing effect, preventing the robotic arm from having an accidental or unstable motion state.

[0046] According to the adaptive control method of the dual-arm robot based on the mirror principle, preferably, in step S6, the weight coefficient adjustment term includes a position weight coefficient K' p , a velocity weight coefficient K' i , an interaction force weight coefficient K' d , and a weight coefficient K' MPC of the prediction model auxiliary force error; each weight coefficient is represented as:

[0047] K′ p =β·K p ;

[0048] K′ i =β·K i ;

[0049] K′d = β · K d ;

[0050] K' MPC = β · K MPC ;

[0051] wherein, K p , K i , K d respectively represent the adaptive weight coefficients of position, velocity, and interaction force, and K MPC represents the prediction model auxiliary force weight coefficient; the weight coefficient adjustment factor β is a fixed amplification coefficient, used to control the amplification or reduction of the overall weight, and the larger the value of β, the higher the sensitivity and response speed of the system; the smaller the value of β, the lower the sensitivity of the system but the higher the stability; K' p adjusts the weight of the position error according to β, used to control the influence of the position error on the auxiliary force; K' i adjusts the weight of the velocity error according to β, used to control the influence of the velocity error on the auxiliary force; K' d adjusts the weight of the interaction force error according to β, used to control the influence of the interaction force error on the auxiliary force; K' MPC adjusts the weight of the prediction model auxiliary force error according to β, used to control the influence of the position error on the auxiliary force.

[0052] Based on the prediction model auxiliary force weight coefficient K MPC and the dynamically adjusted adaptive weight coefficients K' p , K' i , K' d , the minimum auxiliary force required by the patient's affected side mechanical arm is accurately predicted according to the motion state of the healthy side and the affected side and the mechanical arm dynamics characteristics, unnecessary auxiliary force is reduced by introducing the optimization function R(X(t)), and the minimum auxiliary force control for rehabilitation training is ensured, while playing a damping role when the healthy side and the affected side mechanical arms change sharply.

[0053] According to the above-mentioned adaptive control method of the double-arm robot based on the mirror principle, preferably, in step S7, each weight coefficient will be adaptively adjusted according to the difference between the healthy side and the affected side mechanical arms to ensure that the affected side always maintains the minimum auxiliary force exertion.

[0054] The position weight coefficient K' p is used to adjust the influence degree of the position difference on the auxiliary force; the velocity weight coefficient K' i is used to adjust the influence degree of the velocity difference on the auxiliary force; the interaction force weight coefficient K' d is used to adjust the influence degree of the interaction force difference on the auxiliary force; and the prediction model weight coefficient K' MPC is used to adjust the influence degree of the prediction model parameter difference, and the formula is as follows:

[0055] Therefore, Thus,

[0056] Therefore, Thus,

[0057] Therefore, Thus,

[0058] Therefore, Thus,

[0059] wherein Δx represents the position difference between the healthy side and the affected side robot arms, reflecting the difference degree of the position of the healthy side and the affected side robot arms; Δv represents the speed difference between the healthy side and the affected side robot arms, reflecting the difference degree of the speed of the healthy side and the affected side robot arms; ΔF represents the interaction force difference between the healthy side and the affected side robot arms, reflecting the difference degree of the interaction force of the healthy side and the affected side robot arms; and Δτ represents the parameter difference of the predicted model auxiliary force of the healthy side and the affected side robot arms, reflecting the difference degree of the predicted model auxiliary force parameters of the healthy side and the affected side robot arms.

[0060] In the process of patient rehabilitation training, by monitoring the parameter data of the healthy side and the affected side robot arms, i.e. Δx (position difference), Δv (speed difference), ΔF (interaction force difference) and Δτ (predicted model parameter difference), the difference degree between the healthy side and the affected side robot arms is reflected; if the difference of any of these parameters increases, it means that the synchronous movement of the healthy side and the affected side robot arms deviates; at this time, the greater the data parameter difference, the greater the corresponding weight coefficient, and more auxiliary force is needed to maintain their synchronous movement.

[0061] According to the above-mentioned adaptive control method of the dual-arm robot based on the mirror principle, preferably, in step S8, the healthy side robot arm drives the affected side robot arm in the process of mirror rehabilitation training task, and according to the dynamic adjustment of the weight coefficient and the collision coefficient C, the auxiliary force that the affected side robot arm should exert is predicted according to the robot dynamics model and the motion state of the healthy side robot arm;

[0062] The calculation formula of the minimum auxiliary force of the affected side is:

[0063] F assistant (x)=C·(K′ p ·Δx+K′ i ·Δx+K′ d ·ΔF+K′ MPC ·τ MPC )-R(X(t));

[0064] wherein F assistant(x) represents the minimum adaptive assistive force of the affected side robot arm; Delta x represents the position difference between the healthy side and the affected side robot arms, which is used to adjust the assistive force according to the position difference; Delta v represents the velocity difference between the healthy side and the affected side robot arms, which is used to adjust the assistive force according to the velocity difference; Delta F represents the interaction force difference between the healthy side and the affected side robot arms, which is used to adjust the assistive force according to the interaction force difference; Tau MPC represents the model prediction-based assistive force part.

[0065] C represents a collision risk coefficient for controlling the positive and negative directions of the assistive force,

[0066]

[0067] The second aspect of the present application provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements any step of the adaptive control method for a dual-arm robot based on the mirror principle according to the first aspect when executing the computer program.

[0068] The third aspect of the present application provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and wherein the computer program implements any step of the adaptive control method for a dual-arm robot based on the mirror principle according to the first aspect when executed by a computer processor.

[0069] Compared with the prior art, the present application has the following advantages:

[0070] (1) The present application adjusts the adaptive weight coefficient and the prediction assistive force weight coefficient in real time, provides a personalized rehabilitation training scheme according to the actual rehabilitation condition of the patient and the state parameters of the robot arm, and meets the rehabilitation needs of different patients; real-time monitoring of the key parameters of the healthy side and the affected side robot arms, i.e. position, velocity, acceleration and interaction force, can quickly respond to the changes in the patient's actions, timely adjust the size and direction of the assistive force of the healthy side and the affected side robot arms, and ensure fast response and personalized rehabilitation training effect.

[0071] (2) In the assistive force function, the adaptive weight coefficient is combined with the prediction weight coefficient. By introducing an optimization function, the affected side robot arm always provides the minimum necessary assistive force during the mirror rehabilitation training of the patient; at the same time, when the motion of the healthy side and the affected side robot arms changes sharply, the optimization function acts as a damping term to slow down the sudden change of motion and prevent the occurrence of unstable conditions, thereby greatly improving the safety and stability of the rehabilitation training process.

[0072] (3) The application introduces a collision avoidance function in the dual-arm robot system, by discretizing the connecting rod of the mechanical arm into multiple points, and calculating the Euclidean distance between the points of the two mechanical arms in real time; the smallest Euclidean distance is compared with the set safety threshold, and the application can accurately determine whether the mechanical arms are in a collision state. For different situations, the collision coefficient C is dynamically adjusted to change the direction of the auxiliary force, thereby avoiding the collision of the mechanical arms in time; the introduction of this function effectively ensures the safety and stability of the rehabilitation training, and provides a more reliable rehabilitation training environment for the patients.

[0073] (4) The application uses the mirror training task scene to enable the patient to actively participate in the rehabilitation training; specifically, during mirror rehabilitation training, the patient dynamically adjusts the adaptive weight coefficient and the prediction model weight coefficient according to the movement ability difference between the healthy side and the affected side, and adjusts the size of the auxiliary force of the affected side mechanical arm of the patient in the rehabilitation training process in real time. In addition, an optimization function is introduced to ensure that the dual-arm robot always assists the patient's affected arm to move with the smallest auxiliary force during the rehabilitation training process, reduces unnecessary auxiliary force, and plays a damping role when moving sharply, thereby ensuring the safety and practicability of the rehabilitation training, and realizing the individualization and effectiveness of the rehabilitation training. BRIEF DESCRIPTION OF DRAWINGS

[0074] Figure 1 is a schematic diagram of a dual-arm robot;

[0075] Figure 2 is a schematic diagram of a mirror rehabilitation training scene of the application;

[0076] Figure 3 is a flowchart of the application. DETAILED DESCRIPTION

[0077] The application will be further described in detail below through specific embodiments, but the scope of the application is not limited thereto.

[0078] Embodiment 1

[0079] A dual-arm robot adaptive control method based on the mirror principle, as shown in Figure 3 , includes the following steps;

[0080] S1, starting a mirror training task scene such as shooting by using a dual-arm robot, as shown in Figure 2 , the patient is located at an initial position and holds the end handle of the dual-arm robot, as shown in Figure 1 ; according to the shooting task, based on the mirror principle, the healthy side mechanical arm actively drives the affected side mechanical arm to perform rehabilitation training movement; the relevant data of the healthy side and affected side mechanical arms are acquired in real time through joint angle sensors and force sensors, and the parameters of the real-time position x, velocity v, acceleration a, and interaction force F of the TCP of the healthy side and affected side mechanical arms are obtained after processing.

[0081] S2, according to the parameters of the position x, velocity v, acceleration a, interaction force F and the like of the healthy side and the affected side mechanical arms obtained in S1, construct the healthy side mechanical arm state quantity Xhealthy(t) and the affected side mechanical arm state quantity Xaffected(t) at the current moment, obtain the difference between the healthy side mechanical arm and the affected side mechanical arm at the current moment, the healthy side mechanical arm state quantity Xhealthy(t) and the affected side mechanical arm state quantity Xaffected(t) respectively include the state quantity of the position, velocity, acceleration, interaction force and the like of the healthy side and the affected side mechanical arms.

[0082] In step S2,

[0083] The Xhealthy(t) includes the position x d , velocity v d , acceleration a d and interaction force F d and the like of the healthy side mechanical arm at t moment; the healthy side mechanical arm state quantity Xhealthy(t) = (x d , v d , a d , F d ).

[0084] The Xaffected(t) includes the position x m , velocity v m , acceleration a m and interaction force F m and the like of the affected side mechanical arm at t moment; the affected side mechanical arm state quantity Xaffected(t) = (x m , v m , a m , F m ).

[0085] The difference ΔX between the healthy side and the affected side mechanical arms is:

[0086] ΔX = Xhealthy(t) - Xaffected(t);

[0087] ΔX = (x d - x m , v d - v m , a d - a m , F d - F m ).

[0088] S3. Based on the parameters of the healthy and affected robotic arms obtained in S1, input them into the dual-arm robot adaptive control network model. Monitor the changes in parameters such as position x, velocity v, acceleration a, and interaction force F of the healthy and affected robotic arms in real time. Introduce a collision coefficient C to predict the collision risk between the healthy and affected robotic arms. By discretizing the links of each robotic arm into a certain number of points and calculating the Euclidean distance d between each point of the other robotic arm and the current robotic arm point, the minimum value d is selected. min Define the safe distance Z between the healthy and affected robotic arms, and monitor and determine the minimum Euclidean distance d between the healthy and affected robotic arms in real time. min If the distance is less than the safe distance Z, there is a risk of collision. If it is less than the safe distance, set C = -1; otherwise, set C = 1.

[0089] In step S3, to effectively prevent collisions between the healthy and affected robotic arms during training, a collision coefficient C is introduced as follows: First, the links of each robotic arm are discretized into a certain number of points, and the Euclidean distance d between each point of the other robotic arm and the current robotic arm point is calculated. The minimum value d is then selected. min Define the safe distance Z between the healthy and affected robotic arms, and monitor and determine the minimum Euclidean distance d between the healthy and affected robotic arms in real time. min Is it less than the safe distance Z? If d min If ≤ Z, then it is a collision state, and the collision coefficient C = -1; if d min If the value is greater than Z, then it is a non-collision state, and the collision coefficient C = 1, so as to change the direction of the auxiliary force of the robotic arm, thereby effectively avoiding collision between the healthy and affected sides of the robotic arm.

[0090] The Euclidean distance formula and the collision threshold setting formula are as follows:

[0091]

[0092] If d min If the value is less than or equal to Z, it is set as the collision threshold to determine the collision state. The collision coefficient is C = -1, otherwise C = 1.

[0093] S4. Based on the current state variables Xhealthy(t) and Xaffected(t) of the healthy robotic arm, input them into the dual-arm robot adaptive control network model. Use this information to construct dynamic models of the healthy and affected robotic arms, predict the predicted values ​​Nhealthy(t+1) and Naffected(t+1) of the healthy robotic arm at the next moment, and obtain the auxiliary force of the prediction model. The predicted values ​​Nhealthy(t+1) and Naffected(t+1) of the healthy robotic arm at the next moment include the predicted end positions, velocities, accelerations, interaction forces, and other parameters of the healthy and affected robotic arms, respectively.

[0094] In step S4,

[0095] The dynamics model of the healthy-side robot arm is:

[0096]

[0097] wherein M 健 represents the mass matrix of the healthy-side robot arm; represents the end acceleration of the healthy-side robot arm; C 健 represents the damping matrix of the healthy-side robot arm; represents the end velocity of the healthy-side robot arm; K 健 represents the stiffness characteristic of the healthy-side robot arm; x 健 represents the end position of the healthy-side robot arm; F ext represents the force externally acting on the end of the healthy-side robot arm.

[0098] The dynamics model of the healthy-side robot arm is:

[0099]

[0100] wherein M 患 represents the mass matrix of the healthy-side robot arm; represents the end acceleration of the healthy-side robot arm; C 患 represents the damping matrix of the healthy-side robot arm; represents the end velocity of the healthy-side robot arm; K 患 represents the stiffness characteristic of the healthy-side robot arm; x 患 represents the end position of the healthy-side robot arm; F assistant represents the assisting force exerted by the robot on the healthy-side robot arm; F interaction represents the interaction force between the healthy-side robot arm and the external environment.

[0101] The state prediction of the predicted quantity Nhealthy(t+1) of the healthy-side robot arm at the next moment and the predicted quantity Naffected(t+1) of the affected-side robot arm at the next moment is:

[0102]

[0103] wherein X 健 (t) represents the state of the healthy-side robot arm at the current moment; X 患 (t) represents the state of the affected-side robot arm at the current moment; f and g represent functions for predicting the states of the healthy-side and affected-side robot arms, respectively.

[0104] The assisting force τ MPC of the prediction model is obtained as:

[0105]

[0106] wherein N健 N healthy (t+1) represents the acceleration prediction of the healthy side robot arm at the next moment; N 患 N affected (t+1) represents the acceleration prediction of the affected side robot arm at the next moment; N 健 N healthy (t+1) represents the velocity prediction of the healthy side robot arm at the next moment; N 患 N affected (t+1) represents the velocity prediction of the affected side robot arm at the next moment; N 健 N healthy (t+1) represents the displacement prediction of the healthy side robot arm at the next moment; N 患 N affected (t+1) represents the displacement prediction of the affected side robot arm at the next moment; N 患 C represents the mass matrix of the affected side robot arm; C 患 C represents the damping matrix of the affected side robot arm; C 患 C represents the stiffness matrix of the affected side robot arm.

[0107] S5, in the adaptive control strategy of the dual-arm robot, S1-S4 are executed in a loop, and the healthy side robot arm state quantity Xhealthy(t), the affected side robot arm state quantity Xaffected(t), the prediction quantity Nhealthy(t+1) of the healthy side robot arm at the next moment, and the prediction quantity Naffected(t+1) of the affected side robot arm at the next moment are continuously monitored and recorded, an optimization function R(X(t)) is introduced, the optimization function is used to optimize the movement effect of the dual-arm robot and the individualization degree of the rehabilitation training, so as to optimize the control strategy and improve the rehabilitation effect.

[0108] In step S5, during the execution of the rehabilitation training task in a loop, the state quantity X(t) and the prediction quantity N(t) of the patient's healthy side and affected side robot arms are continuously monitored and recorded, and the optimization function R(X(t)) is:

[0109] R(X(t)) = w1·e1 + w2·e2;

[0110] wherein w1 and w2 represent weight coefficients, and w1 + w2 = 1; e1 represents a rehabilitation effect index, e1 > 1, e1 includes individualized setting of joint range of motion size and muscle strength, and the better the rehabilitation effect, the higher the value of e1; e2 represents a safety index, e2 > 1, e2 is the number of times of collision risk warning between the healthy side and the affected side robot arms, and if there is no safety problem in the rehabilitation training process, the value of e2 is higher.

[0111] The optimization function R(X(t)) is used to reduce unnecessary auxiliary force and play a damping role in sharp movement, further optimize the control strategy, and ensure that the patient's affected side robot arm always assists the patient in rehabilitation with the smallest auxiliary force, while the value of R(X(t)) will increase sharply when encountering sudden and rapid mirror training movement;

[0112]

[0113] At this time, the assist force F assistant The commutation plays a damping adjustment and mitigation effect, preventing the robot arm from having an unexpected or unstable motion state.

[0114] S6, circulate S1-S5, based on the obtained multiple sets of data, introduce a weight coefficient adjustment factor β as a fixed scaling coefficient of each weight term, for controlling the amplification or reduction of the overall weight, the value of β is larger, then increase the sensitivity and response speed of the system, the value of β is smaller, then reduce the sensitivity but improve the stability of the system.

[0115] In step S6, the weight coefficient adjustment term includes a position weight coefficient K' p , a speed weight coefficient K' i , an interaction force weight coefficient K' d , and a prediction model auxiliary force error weight coefficient K' MPC ; each weight coefficient is represented as:

[0116] K' p = β·K p ;

[0117] K' i = β·K i ;

[0118] K' d = β·K d ;

[0119] K' MPC = β·K MPC ;

[0120] wherein K p , K i , K d represent the adaptive weight coefficients of position, speed, and interaction force respectively, and K MPC represents the prediction model auxiliary force weight coefficient; the weight coefficient adjustment factor β is a fixed amplification coefficient for controlling the amplification or reduction of the overall weight, the value of β is larger, then increasing the sensitivity and response speed of the system; the value of β is smaller, then reducing the sensitivity but improving the stability of the system; K' p adjusts the weight of the position error according to β, for controlling the influence of the position error on the auxiliary force; K' i adjusts the weight of the speed error according to β, for controlling the influence of the speed error on the auxiliary force; K' d adjusts the weight of the interaction force error according to β, for controlling the influence of the interaction force error on the auxiliary force; K' MPC adjusts the weight of the prediction model auxiliary force error according to β, for controlling the influence of the position error on the auxiliary force.

[0121] Based on the prediction model auxiliary force weight coefficient KMPC and dynamically adjusted adaptive weight coefficient K' p , K' i , K' d According to the motion state and mechanical arm dynamics characteristics of the healthy side and the affected side, the minimum assistance force required by the patient's affected side mechanical arm is accurately predicted, unnecessary assistance force is reduced by introducing the optimization function R(X(t)), and the minimum assistance force control for rehabilitation training is ensured, while playing a damping role when the healthy side and the affected side mechanical arms change sharply.

[0122] S7, cycle S1-S6, by real-time acquisition of the position x, velocity v, acceleration a, interaction force F parameters of the healthy side and the affected side mechanical arm end and the assistance force of the prediction model, according to the difference change of the healthy side and the affected side mechanical arm data, dynamically adjust the adaptive weight coefficient, the prediction model weight coefficient and the optimization function.

[0123] In step S7, each weight coefficient will be adaptively adjusted according to the difference between the healthy side and the affected side mechanical arms to ensure that the affected side always maintains the minimum assistance force.

[0124] Each weight coefficient will be adaptively adjusted according to the difference between the healthy side and the affected side mechanical arm state and prediction quantity parameters, and the optimization function R(X(t)) term is introduced to ensure that the affected side always maintains the minimum assistance force.

[0125] The position weight coefficient K' p is used to adjust the influence degree of position difference on assistance force; the velocity weight coefficient K' i is used to adjust the influence degree of velocity difference on assistance force; the interaction force weight coefficient K' d is used to adjust the influence degree of interaction force difference on assistance force; the prediction model weight coefficient K' MPC is used to adjust the influence degree of prediction model parameter difference, the formula is as follows:

[0126]

[0127] Where, Δx represents the position difference between the healthy side and the affected side mechanical arms, reflecting the difference degree of the healthy side and the affected side mechanical arms position; Δv represents the velocity difference between the healthy side and the affected side mechanical arms, reflecting the difference degree of the healthy side and the affected side mechanical arms velocity; ΔF represents the interaction force difference between the healthy side and the affected side mechanical arms, reflecting the difference degree of the healthy side and the affected side mechanical arms interaction force; Δτ represents the parameter difference of the prediction model assistance force between the healthy side and the affected side mechanical arms, reflecting the difference degree of the prediction model assistance force parameters of the healthy side and the affected side mechanical arms.

[0128] In the process of patient rehabilitation training, the parameter data of the healthy side and the affected side mechanical arm, i.e. Δx (position difference), Δv (velocity difference), ΔF (interaction force difference) and Δτ (predicted model parameter difference), are monitored to reflect the difference between the healthy side and the affected side mechanical arm; if the difference of any of these parameters increases, it means that the synchronous movement of the healthy side and the affected side mechanical arm deviates; at this time, the greater the difference in data parameters, the greater the corresponding weight coefficient, and more assistance force is needed to maintain their synchronous movement.

[0129] S8, according to the dynamic change of each weight coefficient and the optimization function, the minimum assistance force size and direction of the affected side mechanical arm are adjusted in real time to ensure that the patient completes the rehabilitation training task with the minimum assistance force during the rehabilitation training process.

[0130] During the mirror-based rehabilitation training task, the healthy side mechanical arm drives the affected side mechanical arm, and the assistance force that the affected side mechanical arm should exert is predicted according to the dynamic adjustment of the weight coefficient and the collision coefficient C according to the mechanical arm dynamics model and the movement state of the healthy side mechanical arm.

[0131] The calculation formula of the minimum assistance force of the affected side is:

[0132] F assistant (x)=C·(K′ p ·Δx+K′ i ·Δv+K′ d ·ΔF+K′ MPC ·τ MPC )-R(X(t));

[0133] Wherein, F assistant (x) represents the minimum adaptive assistance force of the affected side mechanical arm; Δx represents the position difference between the healthy side and the affected side mechanical arm, which is used to adjust the assistance force according to the position difference; Δv represents the velocity difference between the healthy side and the affected side mechanical arm, which is used to adjust the assistance force according to the velocity difference; ΔF represents the interaction force difference between the healthy side and the affected side mechanical arm, which is used to adjust the assistance force according to the interaction force difference; τ MPC represents the part of the assistance force based on model prediction.

[0134] C represents the collision risk coefficient, which is used to control the positive and negative direction of the assistance force,

[0135]

[0136] R(X(t)) is an optimization function to ensure that the patient completes the rehabilitation training task with the minimum assistance force during the rehabilitation training process.

[0137] S9, after completing each mirror rehabilitation training task, collect and process the data of the healthy side and the affected side of the patient's mechanical arm, and generate a table to record the parameter difference between the healthy side and the affected side of the mechanical arm over time.

[0138] S10, based on the table generated in S9, compare and analyze the size and change trend of the difference between the healthy side and the affected side of the mechanical arm, evaluate the overall recovery of the patient during the rehabilitation process, and determine the effect after the first rehabilitation.

[0139] Embodiment 2

[0140] An electronic device includes a memory and a processor, the memory stores a computer program, and the processor executes the computer program to implement any step of the adaptive control method of the dual-arm robot based on the mirror principle as described in embodiment 1.

[0141] The hardware of the electronic device of the embodiment further includes GPU, display buffer memory, RAMD / A converter and radiator cooperating with the processor; the GPU is responsible for processing the graphic display of the electronic device, provides image rendering and acceleration function, and uses its parallel computing advantage to accelerate processing of large-scale data-intensive tasks such as deep learning training and scientific computing.

[0142] Further, the adaptive control method of the dual-arm robot based on the mirror principle described in embodiment 1 can be implemented as a computer software program. For example, the embodiment includes a computer program product including a computer program carried on a computer readable medium, the computer program including program code for executing the method. In such an embodiment, the computer program can be downloaded and installed from the network, and / or installed from a removable medium. When the computer program is executed by the processor, the above-mentioned functions defined in the method of the present application are performed.

[0143] Embodiment 3

[0144] A computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement any step of the adaptive control method of the dual-arm robot based on the mirror principle as described in embodiment 1.

[0145] The computer-readable medium described in this application may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0146] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Python and C++, as well as conventional procedural programming languages ​​or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0147] In this embodiment, the computer-readable storage medium can be accelerated using hardware such as a GPU. The parallel computing advantage of the GPU is used to accelerate any step in the implementation of a dual-arm robot adaptive control method based on the mirror principle as described in Embodiment 1.

[0148] In summary, the present application effectively overcomes the deficiencies in the prior art, and has a high industrial value. The above examples are intended to illustrate the essential content of the present application, but do not limit the scope of protection of the present application. Those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the essence and protection scope of the technical solutions of the present application.

[0149] The above examples are specific embodiments of the present application, but the embodiments of the present application are not limited by the above examples, and any other combinations, changes, modifications, substitutions, simplifications that do not exceed the design ideas of the present application fall within the protection scope of the present application.

Claims

1. A method for adaptive control of a dual-arm robot based on mirror principle, characterized in that, The method comprises the following steps: S1, starting a shooting mirror image training task scene, the patient is located in the initial position and holds the robot end handle with both arms; According to the shooting task, based on the mirror principle, the healthy side mechanical arm actively drives the affected side mechanical arm to perform rehabilitation training movement; the relevant data of the healthy side and affected side mechanical arms are acquired in real time through joint angle sensors and force sensors, and the real-time position x, velocity v, acceleration a and interactive force F parameters of the healthy side and affected side mechanical arm ends are obtained; S2, according to the parameters obtained in S1, the state quantity Xhealthy(t) of the healthy side mechanical arm and the state quantity Xaffected(t) of the affected side mechanical arm at the current time are constructed, and the difference between the healthy side and affected side mechanical arms at the current time is obtained; S3, according to the parameters of the healthy side and affected side mechanical arms obtained in S1, the changes of the position x, velocity v, acceleration a and interactive force F parameters of the current healthy side and affected side mechanical arms are monitored in real time, and a collision coefficient C is introduced; S4, according to the state quantity Xhealthy(t) of the healthy side mechanical arm and the state quantity Xaffected(t) of the affected side mechanical arm at the current time, the dynamics model of the healthy side and affected side mechanical arms is constructed, the predicted quantity Nhealthy(t+1) of the healthy side mechanical arm at the next time and the predicted quantity Naffected(t+1) of the affected side mechanical arm at the next time are predicted, and the auxiliary force of the prediction model is obtained; In step S4, The dynamics model of the healthy side mechanical arm is: wherein M 健 represents the mass matrix of the healthy-side robot arm; represents the end acceleration of the healthy-side robot arm; C 健 represents the damping matrix of the healthy-side robot arm; represents the end velocity of the healthy-side robot arm; K 健 represents the stiffness characteristic of the healthy-side robot arm; x 健 represents the end position of the healthy-side robot arm; F ext represents the external force acting on the end of the healthy-side robot arm; The dynamics model of the affected side mechanical arm is: wherein M 患 represents the mass matrix of the affected-side robot arm; represents the end acceleration of the affected-side robot arm; C 患 represents the damping matrix of the affected-side robot arm; represents the end velocity of the affected-side robot arm; K 患 represents the stiffness matrix of the affected-side robot arm; x 患 represents the end position of the affected-side robot arm; F assistant represents the assisting force exerted by the robot on the affected-side robot arm; F interaction represents the interaction force between the affected-side robot arm and the external environment; The state prediction of the predicted quantity Nhealthy(t+1) of the healthy side mechanical arm at the next time and the predicted quantity Naffected(t+1) of the affected side mechanical arm at the next time is: where X 健 (t) denotes the state of the healthy-side manipulator at the current time; X 患 (t) denotes the state of the affected-side manipulator at the current time; f and g denote functions for predicting the states of the healthy-side and affected-side manipulators, respectively; The assistance force τ of the prediction model is obtained MPC is represented as: where N 健 (t+1)" represents the acceleration prediction of the healthy-side robot arm at the next time; N 患 (t+1)" represents the acceleration prediction of the affected-side robot arm at the next time; N 健 (t+1)' represents the velocity prediction of the healthy-side robot arm at the next time; N 患 (t+1)' represents the velocity prediction of the affected-side robot arm at the next time; N 健 (t+1) represents the displacement prediction of the healthy-side robot arm at the next time; N 患 (t+1) represents the displacement prediction of the affected-side robot arm at the next time; M 患 represents the mass matrix of the affected-side robot arm; C 患 represents the damping matrix of the affected-side robot arm; K 患 represents the stiffness matrix of the affected-side robot arm; S5, S1-S4 are executed in a loop, the state quantity Xhealthy(t) of the healthy side mechanical arm, the state quantity Xaffected(t) of the affected side mechanical arm, the predicted quantity Nhealthy(t+1) of the healthy side mechanical arm at the next time and the predicted quantity Naffected(t+1) of the affected side mechanical arm at the next time are continuously monitored and recorded, and an optimization function R(X(t)) is introduced; S6, S1-S5 are executed in a loop, based on the obtained multiple groups of data, a weight coefficient adjustment factor β is introduced; S7, S1-S6 are executed in a loop, according to the difference changes of the data of the healthy side and affected side mechanical arms, the adaptive weight coefficient, the prediction model weight coefficient and the optimization function are dynamically adjusted; S8, according to the dynamic changes of each weight coefficient and the optimization function, the size and direction of the minimum auxiliary force of the affected side mechanical arm are adjusted in real time; S9, after completing the mirror rehabilitation training task each time, the data of the healthy side and affected side mechanical arms of the patient are collected and processed, and a table is generated; S10, based on the table generated in S9, the size and change trend of the difference between the healthy side and affected side mechanical arms are compared and analyzed, the overall recovery of the patient in the rehabilitation process is evaluated, and the effect after the first rehabilitation is determined.

2. The dual-arm robot adaptive control method based on the mirror principle of claim 1, wherein, In step S2, The Xhealthy(t) includes the position x of the healthy side mechanical arm at the time t d , the velocity v d , the acceleration a d , and the interaction force F d ; the healthy side mechanical arm state quantity Xhealthy(t) = (x d , v d , a d , F d ); The Xpatient(t) includes the position x of the robotic arm on the affected side at time t. m Speed ​​v m acceleration a m and interaction force F m The state quantity of the affected side robotic arm, X_affected(t), is equal to (x_affected(t)). m ,v m ,a m ,F m ); The difference ΔX between the healthy side and affected side mechanical arms is: ΔX=Xhealthy(t)-Xaffected(t); ΔX = (x d - x m , v d - v m , a d - a m , F d - F m ).

3. The dual-arm robot adaptive control method based on the mirror principle of claim 2, wherein, In step S3, the step of introducing the collision coefficient C is: first discretizing the connecting rod of each mechanical arm into a certain number of points, and calculating the Euclidean distance d between each point of the other side mechanical arm and the current mechanical arm point, and screening out the minimum value d min , defining the safety distance Z between the healthy side and the affected side mechanical arm, and monitoring and judging whether the minimum value d min of the Euclidean distance between the healthy side and the affected side mechanical arm is less than the safety distance Z in real time. The Euclidean distance formula and the collision threshold setting formula are: If d min Z, set as collision threshold, determine collision state, collision coefficient C = -1, otherwise C = 1.

4. The dual-arm robot adaptive control method based on the mirror principle of claim 3, wherein, In step S5, the optimization function is: R(X(t))=w1·e1+w2·e2; Wherein, w1 and w2 represent weight coefficients, and w1+w2=1; e1 represents a rehabilitation effect index, e1>1, e1 includes personalized setting of joint range of motion size, muscle strength, and the better the rehabilitation effect, the higher the value of e1; e2 represents a safety index, e2>1.

5. The dual-arm robot adaptive control method based on the mirror principle of claim 4, wherein, In step S6, the weight coefficient adjustment term includes a position weight coefficient Kp p , a speed weight coefficient Kv i , an interaction force weight coefficient Kf d , and a weight coefficient Kp MPC of a prediction model auxiliary force error, each of which is represented as: K' p = β · K p ; K' i = β · K i ; K' d = β · K d ; K' MPC = β · K MPC ; wherein, K p , K i , K d respectively represent adaptive weight coefficients of position, velocity and interaction force, and K MPC represents a prediction model auxiliary force weight coefficient; the weight coefficient adjustment factor β is a fixed amplification coefficient, used to control the amplification or reduction of the overall weight; The optimization function R(X(t)) is used to reduce unnecessary auxiliary force and play a damping role in sharp movement, further optimize the control strategy, and ensure that the patient's affected side mechanical arm always assists the patient in rehabilitation with the smallest auxiliary force, while the value of R(t(t)) will increase sharply when encountering sudden and rapid mirror training movement; At this time, F assistant The commutation plays a damping adjustment and slowing effect, preventing the robot arm from having an accidental or unstable movement state.

6. The dual-arm robot adaptive control method based on the mirror principle of claim 5, wherein, In step S7, the position weight coefficient K ' p for adjusting the degree of influence of the position difference on the assistance force; the velocity weight coefficient K ' i for adjusting the degree of influence of the velocity difference on the assistance force; the interaction force weight coefficient K ' d for adjusting the degree of influence of the interaction force difference on the assistance force; the prediction model weight coefficient K ' MPC for adjusting the degree of influence of the prediction model parameter difference, as follows: Wherein, Δx represents the position difference between the healthy side and the affected side mechanical arm; Δv represents the speed difference between the healthy side and the affected side mechanical arm; ΔF represents the interaction force difference between the healthy side and the affected side mechanical arm; and Δτ represents the parameter difference of the auxiliary force of the healthy side and the affected side mechanical arm prediction model; In the process of patient rehabilitation training, by monitoring the parameter data of the healthy side and the affected side mechanical arm, i.e. Δx, Δv, ΔF and Δτ, the difference degree between the healthy side and the affected side mechanical arm is reflected.

7. The dual-arm robot adaptive control method based on the mirror principle of claim 6, wherein, In step S8, the healthy side mechanical arm drives the affected side mechanical arm in the process of mirror rehabilitation training task, and according to the dynamic adjustment of the weight coefficient and the collision coefficient C, the auxiliary force that the affected side mechanical arm should exert is predicted according to the mechanical arm dynamics model and the motion state of the healthy side mechanical arm. The calculation formula of the minimum auxiliary force of the affected side is: F assistant (x)=C·(K′ p ·Δx+k′ i ·Δv+K′ d ·ΔF+K′ MPC ·τ MPC )-R(X(t)); where F assistant (x) represents the minimum adaptive assist force of the affected side robot arm; Δx represents the position difference between the healthy side and the affected side robot arms, for adjusting the assist force according to the position difference; Δv represents the speed difference between the healthy side and the affected side robot arms, for adjusting the assist force according to the speed difference; ΔF represents the interaction force difference between the healthy side and the affected side robot arms, for adjusting the assist force according to the interaction force difference; τ MPC represents the assist force part based on model prediction.

8. An electronic device comprising a memory and a processor, said memory having stored thereon a computer program, characterized in that, The processor executes the computer program to realize any step in the adaptive control method of the double-arm robot based on the mirror principle of any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the computer processor to realize any step in the adaptive control method of the double-arm robot based on the mirror principle of any one of claims 1-7.

Images