CN117182929A - Flexible control method and device for on-orbit assembly of double-arm robot - Google Patents

Abstract

The invention relates to the technical field of compliance control of a two-arm space robot, and in particular to a compliance control method and device for on-orbit assembly of a two-arm robot. The method includes: obtaining the current motion state of the two-arm robot; inputting the motion state of the two-arm robot into a pre-trained target model to obtain the desired trajectory of the target object at the current moment and the control of the impedance control model adapted to the current environment Parameters; among them, the target model is obtained by training the preset neural network through the motion state of the dual robotic arms, the operating force state of the dual robotic arms and the motion state of the target object as training samples; based on the desired trajectory of the target object, impedance control The control parameters of the model and the preset dual-loop impedance control model are used to obtain the expected joint angles of the dual-arm robots that are suitable for the current environment, which are used as control instructions for the dual-arm robots to achieve compliant control of the dual-arm robot. The invention can improve the efficiency and flexibility of on-orbit assembly of a two-arm robot.

Description

A compliant control method and device for on-orbit assembly of a two-arm robot

Technical field

The invention relates to the technical field of compliance control of a two-arm space robot, and in particular to a compliance control method and device for on-orbit assembly of a two-arm robot.

Background technique

With the widespread application of ground robot technology and the rapid development of aerospace technology, the application of robots in space on-orbit services has shown great advantages and benefits. Compared with the single-arm robot system, the collaborative operation of the two-arm robot allows it to handle more diverse operating tasks. At the same time, the system has a strong load capacity and is suitable for the operation of large inertia objects or flexible objects. Therefore, the dual-arm robot is used in Space on-orbit services are of great significance.

At present, simple motion planning and position control cannot achieve dual-arm cooperative operations involving force and position coordination. Therefore, a compliant control method is needed to ensure that the target object does not slip or cause damage to the system during the entire on-orbit service of the dual-arm robot. Irreversible damage.

In related technologies, compliance control methods all use a framework in which motion planning and impedance controllers are separated. When the position of the target object changes, the path often needs to be re-planned, resulting in inefficient on-orbit assembly; in addition, the impedance controller in this framework The parameters remain basically unchanged during the operation, and the algorithm has poor adaptability and robustness, making the dual-arm operation less flexible.

Based on this, there is an urgent need for a compliance control method and device for on-orbit assembly of a two-arm robot to solve the technical problems of low efficiency and poor flexibility during on-orbit assembly of a two-arm robot.

Contents of the invention

In order to solve the technical problems of low efficiency and poor flexibility during on-orbit assembly of a two-arm robot, embodiments of the present invention provide a compliance control method and device for on-orbit assembly of a two-arm robot.

In the first aspect, embodiments of this specification provide a compliant control method for on-orbit assembly of a dual-arm robot, including:

Get the current motion status of the dual-arm robot;

Input the current motion state of the two-arm robot into the pre-trained target model to obtain the desired trajectory of the current target object and the control parameters of the impedance control model adapted to the current environment; wherein, the target model is obtained by The motion status of the robotic arm, the operating force status of the dual robotic arms and the motion status of the target object are used as training samples to train the preset neural network;

Based on the expected trajectory of the current target object, the control parameters of the impedance control model and the preset dual-loop impedance control model, the expected joint angles of the dual manipulators that are suitable for the current environment are obtained as the control instructions for the dual manipulators. , to achieve compliant control of the two-arm robot.

In a second aspect, embodiments of the present invention also provide a compliance control device for on-orbit assembly of a dual-arm robot, including:

Acquisition module, used to obtain the current motion status of the dual-arm robot;

An input module for inputting the current motion state of the two-arm robot into a pre-trained target model to obtain the desired trajectory of the current target object and the control parameters of the impedance control model adapted to the current environment;

The calculation module is used to obtain the expected joint angles of the dual manipulators that are adapted to the current environment based on the expected trajectory of the current target object, the control parameters of the impedance control model, and the preset dual-loop impedance control model to achieve the dual-loop impedance control model. Compliance control of arm robots.

In a third aspect, embodiments of this specification also provide an electronic device, including a memory and a processor. A computer program is stored in the memory. When the processor executes the computer program, it implements any of the embodiments of this specification. method described.

In a fourth aspect, embodiments of this specification also provide a computer-readable storage medium on which a computer program is stored. When the computer program is executed in a computer, the computer is caused to execute the method described in any embodiment of this specification. .

Embodiments of this specification provide a compliant control method and device for on-orbit assembly of a two-arm robot. By inputting the obtained current motion state of the two-arm robot into a pre-trained target model, the desired trajectory of the current target object is obtained. and the control parameters of the impedance model that are suitable for the current environment. Finally, based on the expected trajectory of the current target object, the control parameters of the impedance model, and the preset double-loop impedance control model, the expected joint angles of the dual manipulators that are suitable for the current environment are obtained. , used as the control command of the dual-arm robot to achieve compliant control of the dual-arm robot. Therefore, the above solution can realize the simultaneous optimization of motion planning and impedance control model control parameters of the robot's dual manipulators, and can adaptively adjust the parameters of the impedance controller according to the current environmental characteristics, thus improving the on-orbit assembly efficiency of the dual-arm robot. Efficiency and flexibility of dual-arm robot operation.

Description of the drawings

In order to more clearly explain the embodiments of this specification or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are: For some embodiments of this specification, those of ordinary skill in the art can also obtain other drawings based on these drawings without exerting creative efforts.

Figure 1 is a flow chart of a compliance control method for on-orbit assembly of a dual-arm robot provided by an embodiment of this specification;

Figure 2 is a hardware architecture diagram of an electronic device provided by an embodiment of this specification;

Figure 3 is a structural diagram of a compliance control device for on-orbit assembly of a two-arm robot provided by an embodiment of this specification.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of this specification clearer, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments It is a part of the embodiments in this specification, not all the embodiments. Based on the embodiments in this specification, all other embodiments obtained by those of ordinary skill in the art without any creative work are protected by this specification. scope.

Please refer to Figure 1. The embodiment of this specification provides a compliant control method for on-orbit assembly of a dual-arm robot. The method includes:

Step 100: Obtain the current motion status of the dual-arm robot;

Step 102: Input the current motion state of the two-arm robot into the pre-trained target model to obtain the desired trajectory of the current target object and the control parameters of the impedance control model adapted to the current environment; wherein, the target model It is obtained by training the preset neural network through the motion state of the dual robotic arms, the operating force state of the dual robotic arms and the motion state of the target object as training samples;

Step 104: Based on the expected trajectory of the current target object, the control parameters of the impedance control model and the preset double-loop impedance control model, obtain the expected joint angles of the dual manipulators that are suitable for the current environment as the control instructions for the dual manipulators. Achieve compliant control of dual-arm robots.

In this embodiment, by inputting the obtained current motion state of the two-arm robot into the pre-trained target model, the desired trajectory of the target object and the control parameters of the impedance model adapted to the current environment are obtained. Finally, based on the target The expected trajectory of the object, the control parameters of the impedance model and the preset double-loop impedance control model are used to obtain the expected joint angles of the dual-arms that are suitable for the current environment, which are used as control instructions for the dual-arms to achieve compliant control of the dual-arm robot. . Therefore, the above solution can realize the simultaneous optimization of motion planning and impedance control model control parameters of the robot's dual manipulators, and can adaptively adjust the parameters of the impedance controller according to the current environmental characteristics, thereby improving the on-orbit assembly efficiency of the dual-arm robot. and flexibility in dual-arm robot operation.

The execution of each step shown in Figure 1 is described below.

For step 100, obtain the current motion state of the dual-arm robot.

It should be noted that in this embodiment, the motion state of the two-arm robot at the current moment specifically refers to the posture and speed of the end effectors of the dual manipulator arms of the dual-arm robot at the current moment.

For step 102:

In one embodiment of this specification, the motion state of the dual robotic arms includes the joint angles, joint angular velocities, and posture and speed of the end effectors of the dual robotic arms;

The operating force state of the dual robotic arms includes six-dimensional force information at the ends of the dual robotic arms and the environmental external force on the center of mass of the target object;

The motion state of the target object includes the pose of the target object's center of mass.

In this implementation, in order to obtain the desired motion of the target object at the current moment and the impedance controller parameter adjustment strategy adapted to the current environment, the above-obtained motion status of the dual manipulator arms, the operating force information of the dual manipulator arms and the target object's The motion state is used as a training sample, and online training of target object motion planning and impedance controller parameter adjustment strategies is carried out through the preset neural network until the algorithm converges; in order to improve learning efficiency, the error vector between the object's current position and the target position is also As state information of training samples.

In one embodiment of this specification, the preset neural network includes a LSTM network module, an MLP network module, a deep reinforcement learning network module and an experience replay pool module that are connected in sequence.

Considering that the deep reinforcement learning network algorithm has strong generalization characteristics and can improve the autonomy of system planning and control, in this embodiment, based on the deep reinforcement learning network, the LSTM network, MLP network, deep reinforcement learning network and experience replay pool are combined It is designed as a centralized neural network structure, thereby unifying the adaptive adjustment of the control parameters of the dual manipulator motion planning and the impedance control model under the same framework, thus solving the dual problems caused by the separation of the manipulator motion planning and the impedance control model in related technologies. The on-orbit assembly efficiency of the robotic arm is low, and the poor adaptability and robustness of the traditional impedance control model result in poor flexibility of the dual robotic arms.

In one embodiment of this specification, the target model is trained in the following manner:

Obtain the environmental state quantity according to the preset frequency; where the environmental state quantity includes the operating force state of the dual robotic arms, the motion state of the dual robotic arms, and the motion state of the target object;

Input the operating force status of the dual robotic arms into the LSTM network module to obtain the operating force status of the dual robotic arms with time series;

Use the MLP network module to extract features from the operating force status of the dual robotic arms, and obtain the feature vector of the force status of the dual robotic arms at the current moment;

Use the feature vector of the force state of the dual robotic arms, the motion state of the dual robotic arms and the motion state of the target object as the input state vector of the deep reinforcement learning network module;

Set the environmental reward function of the agent, use the deep reinforcement learning network module to train the agent, and obtain the output action of the target object; the output action of the target object includes the position change of the center of mass of the target object and the control parameters of the impedance control model;

Store the obtained environmental state quantity, the output action of the target object, the environmental reward value obtained by executing the output action, and the environmental state quantity after executing the action into the experience playback pool module to obtain experience data;

Use empirical data to update the current neural network until the algorithm enters a convergence state and obtain the target model.

In this embodiment, the motion state of the dual manipulator arms and the motion state of the target object are obtained according to a preset frequency of 20 Hz. This frequency is beneficial to ensuring the stability of the operation of the dual manipulator arms. At the same time, it is assumed that the control frequency of the impedance control model is h, then it is assumed that the operating force state of the dual robotic arms obtained by the neural network is a time series with a length of h/20. At each time t, the operating force state of the dual robotic arms with a shape of [h/20,3] is The time series is input into the LSTM network module, and then the LSTM network module outputs the time series of the operating force state with the same shape (where 3 indicates that the operating force state of the double manipulator is a force in the three-dimensional direction); after that, a layer of MLP network is used The module performs feature extraction on the time series of operating force states output from the LSTM network module to obtain the feature vector of the operating force state at the current moment; finally, the motion state of the robotic arm, the motion state of the target object, and the current moment are obtained at the current moment. The feature vectors of the operating force states are connected so that they jointly serve as the input state vector of the deep reinforcement learning network.

In this embodiment, considering that the dual manipulators of the two-arm robot have strong timing when cooperatively operating the same target object, and the traditional reinforcement learning network cannot handle the problem of high-dimensional timing information, a network with LSTM is designed. The centralized neural network structure of the module and the MLP network module. This centralized network structure can synchronously obtain the motion status and external interaction force status of the dual robotic arms, thereby realizing the simultaneous optimization of the motion planning and impedance control model parameters of the dual robotic arms of the robot. This allows the two-arm robot to adaptively adjust the control parameters of the impedance control model according to the current environmental characteristics and system motion status during operation, thereby improving the efficiency of the on-orbit assembly of the two-arm robot and the flexibility of the two-arm operation.

At the same time, in this embodiment, the LSTM network module is first used to obtain the high-dimensional time-series force state information of the dual robotic arms obtained at each moment, and then the MLP network module is used to extract features of the time-series force state information output by the LSTM network module. The force trend feature in the vector significantly improves the compliance operation performance.

For step 104:

In one embodiment of this description, step 104 may specifically include the following steps:

According to the desired trajectory of the target object, the kinematic constraints and dynamic model of the closed-chain system, the desired trajectory and desired operating force at the end of the dual manipulator are obtained;

According to the control parameters of the impedance control model, the preset double-loop control model and the expected trajectory and expected operating force of the end of the double manipulator, the compliant expected trajectory of the end of the double manipulator is obtained;

According to the compliant desired trajectory of the end of the dual robotic arm and the inverse kinematics formula of the robotic arm, the desired joint angle of the dual robotic arm is obtained.

In this embodiment, the target model obtained after training is applied to the robot's dual-arm collaborative assembly operation. The trained target model obtains the position change of the center of mass of the current target object by acquiring the motion state of the dual-arm robot at the current moment. (i.e., the desired trajectory of the target object) and the control parameters of the impedance control model adapted to the current environment. In the impedance control model, the expected inertia has a greater impact on the stability of the entire dual-arm robot system, but its impact on the final motion state of the dual-arm robot is small. During the training process and the entire on-orbit assembly process, this implementation Example sets the desired inertia to a fixed parameter, which can be 1, for example. Therefore, in this embodiment, the control parameters of the impedance control model output by the target model are damping parameters and stiffness parameters that are suitable for the current environment.

In one embodiment of this specification, the preset dual-loop impedance control model includes an outer-loop impedance model and an inner-loop impedance model; where the outer-loop impedance model is used to correct the desired trajectory of the target object, and the inner-loop impedance model is used to correct the dual-machine impedance model. The desired trajectory of the arm;

First, establish a second-order impedance model in Cartesian space:

In the formula, ,/> and/> are the expected inertia, damping and stiffness of the second-order impedance model, /> ,/> and/> are the actual acceleration, speed and position of the robotic arm in Cartesian space,/> ,/> and/> are the expected acceleration, speed and position of the robotic arm respectively,/> is the actual contact force,/> is the expected contact force of the robotic arm. In this embodiment, the three impedance parameters are all positive definite matrices. A diagonal matrix is usually selected to obtain a linear decoupling response;

An outer loop impedance model is established between the target object and the external environment, and the second-order impedance model in Cartesian space is rewritten to obtain the outer loop impedance control model:

,/> ,/>

In the formula, Is the actual trajectory of the target object/> and expected trajectory/> The trajectory error between is the first derivative of the target object trajectory error,/> is the second derivative of the target object trajectory error,/> It is the external force exerted by the environment on the target object,/> ,/> , and/> are the expected inertia, damping and stiffness of the outer loop impedance model,/> is the actual trajectory of the target object/> The first derivative of ,/> is the expected trajectory of the target object/> The first derivative of ,/> is the actual trajectory of the target object/> The second derivative of ,/> as target

The actual trajectory of the object The second derivative of ;

Then the inner loop impedance control model of the end operating force of the robot arm i is established. The inner loop impedance model is:

,/> ,/> ,/>

In the formula, is the actual trajectory of robotic arm i/> and expected trajectory/> The trajectory error between Expected force at end of robotic arm i/> and actual operation ability/> The error between ,/> and/> are the expected inertia, damping and stiffness of the inner loop impedance model, /> is the actual trajectory of robotic arm i/> The first derivative of ,/> is the expected trajectory of robotic arm i/> The first derivative of ,/> is the actual trajectory of robotic arm i/> The second derivative of ,/> is the expected trajectory of robotic arm i/> the second derivative of .

In this embodiment, the preset outer loop impedance control model is used to correct the expected trajectory of the current target object. When there is no interaction between the target object and the environment, the interference force of the environment on the target object is minimized. The outer loop impedance The model no longer corrects the desired trajectory of the target object, but obtains the compliant desired trajectory of the target object, thereby achieving compliant control between the target object and the environment.

In one embodiment of this specification, during the coordinated operation of the dual robotic arms, a stable rigid connection needs to be maintained between the dual robotic arms and the target object. Therefore, the actuator at the end of the dual robotic arms and the target object need to be closed at all times. Chain kinematics constraints, the kinematics constraints of the closed chain system are:

In the formula, is the homogeneous transformation matrix of the end-effector coordinate system of robotic arm i relative to its base coordinate system,/> is the homogeneous transformation matrix of the world coordinate system relative to the target object center of mass coordinate system, /> It can be obtained from the pose of the contact point between the end effector of the robotic arm i and the target object, /> Obtained from the inversion of the base coordinate system of robot arm i;

The dynamic model includes the force balance equation and the moment balance equation of the target object; among them, the force balance equation of the target object is:

The moment balance equation of the target object is:

In the formula, and/> are the external forces and moments exerted by the environment on the target object,/> ,/> ,/> and/> are the force and torque exerted on the target object by the end effectors of the left and right manipulator arms respectively,/> ,/> ,/> They are/> , ,/> The vector of the force on the target object acting on the center of mass of the object,/> and/> are the linear velocity and angular velocity of the center of mass of the target object,/> is the mass of the target object,/> Moment of inertia of the target object,/> is the gravity of the target object.

In this embodiment, first, the kinematic constraints of the above-mentioned closed-chain system are used to decompose the compliant expected trajectory of the current target object. Specifically, first, the target object's center of mass relative to the world coordinates at each moment is obtained from the compliant trajectory of the target object. The transformation matrix of the system , find the inverse matrix of this matrix and substitute it into the kinematic constraints of the above closed-chain system , calculate the transformation matrix of the manipulator end effector relative to the base coordinate system/> , the expected trajectory of the double-arm end effector can be obtained. According to the above dynamic model, the expected operating force of the double-arm end effector is calculated. After that, the control parameters of the impedance control model output by the target model that are suitable for the current environment and The expected trajectory and expected operating force at the end of the dual manipulator calculated above are substituted into the above inner loop impedance control model, and the compliant trajectory at the end of the dual manipulator is calculated. Finally, the inverse kinematics formula of the manipulator is calculated to obtain the corresponding compliant trajectory. The desired joint angles of the dual robotic arms are used as control instructions for the dual robotic arms to achieve compliant control of the robot's on-orbit assembly.

In summary, this embodiment provides a compliance control method and device for on-orbit assembly of a dual-arm robot, which uses a centralized neural network structure to unify the motion planning of the target object and the adjustment of parameters of the impedance control model. The centralized neural network structure The network structure can synchronously obtain the motion status and interaction force status of the dual robotic arms, so that the control parameters of the impedance control model can be adaptively adjusted according to the current environmental characteristics and system motion status, thus improving the efficiency of the on-orbit assembly of the dual robotic arms and the dual-arm assembly efficiency. Flexibility of robotic arm operation. At the same time, the LSTM network module and the MLP network module are used to pre-extract features from the force state information during operation. The force trend characteristics contained in the feature vector significantly improve the compliance operation performance.

As shown in Figures 2 and 3, embodiments of this specification provide a compliance control device for on-orbit assembly of a two-arm robot. The device embodiments may be implemented by software, or may be implemented by hardware or a combination of software and hardware. From the hardware level, as shown in Figure 2, it is a hardware architecture diagram of the electronic equipment in which the compliance control device of a two-arm robot assembled in orbit provided by the embodiment of this specification. In addition to the processor shown in Figure 2, In addition to memory, network interfaces, and non-volatile storage, the electronic device where the device in the embodiment is located may also generally include other hardware, such as a forwarding chip responsible for processing messages, etc. Taking software implementation as an example, as shown in Figure 3, as a logical device, it is formed by reading the corresponding computer program in the non-volatile memory into the memory and running it through the CPU of the electronic device where it is located.

As shown in Figure 3, this embodiment provides a compliance control device for on-orbit assembly of a two-arm robot, including:

The acquisition module 300 is used to acquire the current motion state of the dual-arm robot;

The input module 302 is used to input the motion state of the two-arm robot into the pre-trained target model to obtain the desired trajectory of the target object and the control parameters of the impedance control model that are suitable for the current environment;

The calculation module 304 is used to obtain the expected joint angle of the dual-arm robot that is suitable for the current environment based on the expected trajectory of the target object, the control parameters of the impedance control model, and the preset dual-loop impedance control model, so as to achieve the compliance of the dual-arm robot. control.

In this embodiment of the present description, the acquisition module 300 can be used to perform step 100 in the above method embodiment, the input module 302 can be used to perform step 102 in the above method embodiment, and the calculation module 304 can be used to perform the steps in the above method embodiment. 104.

In one embodiment of this specification, the motion state of the dual robotic arms includes the joint angles, joint angular velocities of the dual robotic arms, and the posture and speed of the end effectors of the dual robotic arms;

The operating force state of the dual robotic arms includes six-dimensional force information at the ends of the dual robotic arms and the environmental external force experienced by the center of mass of the target object;

The motion state of the target object includes the posture of the center of mass of the target object.

In one embodiment of this specification, the preset neural network includes an LSTM network module, an MLP network module, a deep reinforcement learning network module and an experience replay pool module that are connected in sequence.

In one embodiment of this specification, the target model is trained in the following manner:

Obtain the environmental state quantity according to a preset frequency; wherein the environmental state quantity includes the operating force state of the dual robotic arms, the motion state of the dual robotic arms, and the motion state of the target object;

Input the operating force status of the dual robotic arms into the LSTM network module to obtain the operating force status of the dual robotic arms with time series;

Use the MLP network module to perform feature extraction on the operating force state of the dual robotic arms to obtain the feature vector of the force state of the dual robotic arms at the current moment;

Use the feature vector of the force state of the dual robotic arms, the motion state of the dual robotic arms and the motion state of the target object as the input state vector of the deep reinforcement learning network module;

Set the environmental reward function of the agent, use the deep reinforcement learning network module to train the agent, and obtain the output action of the target object; wherein the output action of the target object includes the position change of the center of mass of the target object and the control of the impedance control model parameter;

Store the obtained environmental state quantity, the output action of the target object, the environmental reward value obtained by executing the action, and the environmental state quantity after executing the action into the experience playback pool module to obtain experience data;

The current neural network is updated using the empirical data until the algorithm enters a convergence state and the target model is obtained.

In one embodiment of this specification, the preset dual-loop impedance model includes an outer loop impedance model and an inner loop impedance model; where,

The outer loop impedance model is:

,/> ,/>

In the formula, is the trajectory error between the actual trajectory and the expected trajectory of the target object,/> is the first derivative of the target object trajectory error,/> is the second derivative of the target object trajectory error,/> It is the external force exerted by the environment on the target object,/> ,/> and/> are the expected inertia, damping and stiffness of the outer loop impedance model respectively;

The inner loop impedance model is:

,/> ,/>

In the formula, is the trajectory error between the actual trajectory and the desired trajectory of robotic arm i,/> The error between the expected force at the end of robot arm i and the actual operating force,/> ,/> and/> are the expected inertia, damping and stiffness of the inner loop impedance model, respectively.

In one embodiment of this specification, the computing module is used to perform the following operations:

According to the desired trajectory of the target object, the kinematic constraints and dynamic model of the closed-chain system, the desired trajectory and desired operating force of the end of the dual manipulator are obtained;

According to the control parameters of the impedance control model, the preset double-loop control model and the expected trajectory and expected operating force of the end of the dual manipulator, a compliant desired trajectory of the end of the dual manipulator is obtained;

According to the compliant desired trajectory of the end of the dual robotic arm and the inverse kinematics formula of the robotic arm, the desired joint angle of the dual robotic arm is obtained.

In one embodiment of this specification, the kinematic constraint of the closed-chain system is:

In the formula, is the homogeneous transformation matrix of the end-effector coordinate system of robotic arm i relative to its base coordinate system,/> is the homogeneous transformation matrix of the world coordinate system relative to the target object center of mass coordinate system, /> It can be obtained from the pose of the contact point between the end effector of the robotic arm i and the target object, /> Obtained from the inversion of the base coordinate system of robot arm i;

The dynamic model includes the force balance equation and the moment balance equation of the target object; wherein, the force balance equation of the target object is:

The moment balance equation of the target object is:

In the formula, and/> are the external forces and moments exerted by the environment on the target object,/> ,/> ,/> and/> They are the force and torque exerted on the target object by the end effectors of the left and right manipulators/> ,/> ,/> They are/> ,/> , The vector of the force on the target object acting on the center of mass of the object,/> and/> are the linear velocity and angular velocity of the center of mass of the target object,/> is the mass of the target object,/> Moment of inertia of the target object,/> is the gravity of the target object.

It can be understood that the structures illustrated in the embodiments of this specification do not constitute a specific limitation on a compliance control device assembled on-orbit for a two-arm robot. In other embodiments of this specification, a spacecraft evasive maneuver control device may include more or less components than shown in the figure, or combine some components, or separate some components, or arrange different components. . The components illustrated may be implemented in hardware, software, or a combination of software and hardware.

The information interaction, execution process, etc. between the modules in the above device are based on the same concept as the method embodiments of this specification. For specific content, please refer to the description in the method embodiments of this specification, and will not be described again here.

Embodiments of this specification also provide an electronic device, including a memory and a processor. A computer program is stored in the memory. When the processor executes the computer program, a dual function in any embodiment of this specification is implemented. Compliance control method for on-orbit assembly of arm robots.

Embodiments of this specification also provide a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, it causes the processor to execute any implementation of this specification. An example of a compliant control method for on-orbit assembly of a two-arm robot.

Specifically, a system or device equipped with a storage medium may be provided, on which the software program code that implements the functions of any of the above embodiments is stored, and the computer (or CPU or MPU) of the system or device ) reads and executes the program code stored in the storage medium.

In this case, the program code itself read from the storage medium can implement the functions of any one of the above embodiments, and therefore the program code and the storage medium storing the program code form a part of this specification.

Examples of storage media for providing program codes include floppy disks, hard disks, magneto-optical disks, optical disks (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), Tapes, non-volatile memory cards and ROM. Alternatively, the program code can be downloaded from the server computer via the communications network.

In addition, it should be clear that the above embodiments can be implemented not only by executing the program code read by the computer, but also by causing the operating system etc. operating on the computer to complete some or all of the actual operations through instructions based on the program code. function of any embodiment.

In addition, it can be understood that the program code read from the storage medium is written into a memory provided in an expansion board inserted into the computer or into a memory provided in an expansion module connected to the computer, and then the program code is read based on the program code. The instructions cause the CPU installed on the expansion board or expansion module to perform part or all of the actual operations, thereby realizing the functions of any of the above embodiments.

It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is a relationship between these entities or operations. There is no such actual relationship or sequence. Furthermore, the terms "comprises," "comprises," or any other variations thereof are intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus that includes a list of elements includes not only those elements, but also those not expressly listed other elements, or elements inherent to the process, method, article or equipment.

Those of ordinary skill in the art can understand that all or part of the steps to implement the above method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, It includes the steps of the above method embodiment; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk and other various media that can store program codes.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present specification, but not to limit it; although the present specification has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that it can still be used Modifications may be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions may be made to some of the technical features; however, these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions in the embodiments of this specification.

Claims (10)

1. A compliant control method for on-orbit assembly of a two-arm robot, characterized by including: Get the current motion status of the dual-arm robot; Input the current motion state of the two-arm robot into the pre-trained target model to obtain the desired trajectory of the current target object and the control parameters of the impedance control model adapted to the current environment; wherein, the target model is obtained by The motion status of the robotic arm, the operating force status of the dual robotic arms and the motion status of the target object are used as training samples to train the preset neural network; Based on the expected trajectory of the current target object, the control parameters of the impedance control model and the preset dual-loop impedance control model, the expected joint angles of the dual manipulators that are suitable for the current environment are obtained as the control instructions for the dual manipulators. , to achieve compliant control of the two-arm robot. 2. The method according to claim 1, characterized in that the motion state of the dual robotic arms includes joint angles, joint angular velocities of the dual robotic arms, and posture and speed of the end effectors of the dual robotic arms; The operating force state of the dual robotic arms includes six-dimensional force information at the ends of the dual robotic arms and the environmental external force experienced by the center of mass of the target object; The motion state of the target object includes the posture of the center of mass of the target object. 3. The method according to claim 1, wherein the preset neural network includes an LSTM network module, an MLP network module, a deep reinforcement learning network module and an experience replay pool module connected in sequence. 4. The method according to claim 3, characterized in that the target model is trained in the following manner: Obtain the environmental state quantity according to a preset frequency; wherein the environmental state quantity includes the operating force state of the dual robotic arms, the motion state of the dual robotic arms, and the motion state of the target object; Input the operating force status of the dual robotic arms into the LSTM network module to obtain the operating force status of the dual robotic arms with time series; Use the MLP network module to perform feature extraction on the operating force state of the dual robotic arms to obtain the feature vector of the force state of the dual robotic arms at the current moment; Use the feature vector of the force state of the dual robotic arms, the motion state of the dual robotic arms and the motion state of the target object as the input state vector of the deep reinforcement learning network module; Set the environmental reward function of the agent, use the deep reinforcement learning network module to train the agent, and obtain the output action of the target object; wherein the output action of the target object includes the position change of the center of mass of the target object and the change of the impedance control model Control parameters; Store the obtained environmental state quantity, the output action of the target object, the environmental reward value obtained by executing the action, and the environmental state quantity after executing the action into the experience playback pool module to obtain experience data; The current neural network is updated using the empirical data until the algorithm enters a convergence state and the target model is obtained. 5. The method according to claim 1, wherein the preset dual-loop impedance model includes an outer loop impedance model and an inner loop impedance model; wherein, the outer loop impedance model is: ,/> ,/> In the formula, is the trajectory error between the actual trajectory and the expected trajectory of the target object,/> is the first derivative of the target object trajectory error,/> is the second derivative of the target object trajectory error,/> It is the external force exerted by the environment on the target object,/> ,/> and/> are the expected inertia, damping and stiffness of the outer loop impedance model respectively; The inner loop impedance model is: ,/> ,/> In the formula, is the trajectory error between the actual trajectory and the desired trajectory of robotic arm i,/> The error between the expected force at the end of robot arm i and the actual operating force,/> ,/> and/> are the expected inertia, damping and stiffness of the inner loop impedance model, respectively. 6. The method according to claim 1, characterized in that, based on the expected trajectory of the current target object, the control parameters of the impedance control model and the preset dual-loop impedance control model, a dual-loop impedance control model adapted to the current environment is obtained. Desired joint angles of the robotic arm, including: According to the desired trajectory of the current target object, the kinematic constraints and dynamic model of the closed-chain system, the desired trajectory and desired operating force of the end of the dual manipulator are obtained; According to the control parameters of the impedance control model, the preset double-loop control model and the expected trajectory and expected operating force of the end of the dual manipulator, a compliant desired trajectory of the end of the dual manipulator is obtained; According to the compliant desired trajectory of the end of the dual robotic arm and the inverse kinematics formula of the robotic arm, the desired joint angle of the dual robotic arm is obtained. 7. The method according to claim 6, characterized in that, The kinematic constraints of the closed-chain system are: In the formula, is the homogeneous transformation matrix of the end-effector coordinate system of robotic arm i relative to its base coordinate system,/> is the homogeneous transformation matrix of the world coordinate system relative to the target object center of mass coordinate system, /> It can be obtained from the pose of the contact point between the end effector of the robotic arm i and the target object, /> Obtained from the inversion of the base coordinate system of robot arm i; The dynamic model includes the force balance equation and the moment balance equation of the target object; wherein, the force balance equation of the target object is: The moment balance equation of the target object is: In the formula, and/> are the external forces and moments exerted by the environment on the target object,/> , /> ,/> and/> are the force and torque exerted on the target object by the end effectors of the left and right manipulator arms respectively,/> ,/> ,/> They are/> ,/> ,/> The vector of the force on the target object acting on the center of mass of the object,/> and/> are the linear velocity and angular velocity of the center of mass of the target object,/> is the mass of the target object,/> Moment of inertia of the target object,/> is the gravity of the target object. 8. A compliance control device for on-orbit assembly of a two-arm robot, characterized by including: Acquisition module, used to obtain the current motion status of the dual-arm robot; An input module for inputting the current motion state of the two-arm robot into a pre-trained target model to obtain the desired trajectory of the target object at the current moment and the control parameters of the impedance control model adapted to the current environment; The calculation module is used to obtain the expected joint angle of the dual manipulator adapted to the current environment based on the expected trajectory of the target object at the current moment, the control parameters of the impedance control model and the preset double-loop impedance control model, as the The control instructions of the dual-arm robot realize the compliant control of the dual-arm robot. 9. An electronic device, including a memory and a processor. A computer program is stored in the memory. When the processor executes the computer program, the method according to any one of claims 1-7 is implemented. 10. A computer-readable storage medium having a computer program stored thereon, which when the computer program is executed in a computer, causes the computer to perform the method according to any one of claims 1-7.