A teleoperation skill transfer method combined with policy adjustment

Abstract

The application discloses a teleoperation skill transfer method and device combined with strategy adjustment, a medium and equipment. The method comprises the following steps: collecting demonstration data by completing an operation task based on a pre-constructed force feedback teleoperation system; determining a main motion direction according to interactive force data and contact force data; dividing each stage of the operation task based on the main motion direction, so that each task stage has corresponding demonstration data; determining a strategy adjustment parameter in each task stage based on the demonstration data; using a dynamic motion primitive model to stage-coding a robot running track, and calculating a mapping relationship between the strategy adjustment parameter and the running track in the main motion direction through an interpolation algorithm; and generating a target action of the robot based on a pre-constructed motion planner, so that the robot can autonomously complete a contact task. The application realizes that the action planning strategy of a human in a contact interaction task is included in human-machine skill transfer, and the robot can learn the strategy adjustment capability of the human in the task.

Description

Technical Field

[0001] This application relates to the field of robot control, and more particularly to a method, apparatus, medium, and device for transferring teleoperation skills incorporating strategy adjustments. Background Technology

[0002] Currently, robots can replace humans in performing diverse tasks in industrial fields or hazardous areas. Through force feedback teleoperation, operators can remotely control robots by sensing the robot's interaction with the environment or target objects and making real-time decisions and reactions, thereby extending human behavioral and perceptual capabilities. When interacting with unknown environments, human operators can adaptively adjust their motion strategies based on task requirements and environmental perception during interactions with force fields or obstacles, ensuring the arm exhibits appropriate stiffness and limiting and balancing trajectory deviations and contact forces during contact. However, when the robot autonomously performs contact operations, or when the teleoperation system malfunctions, the robot needs to adjust its movements and trajectory planning strategies in real-time during interactions with unknown environments. Traditional robot trajectory tracking and impedance control suffer from low flexibility and poor adaptability in such scenarios. Variable impedance control is commonly used to address the scenarios described above, but the design and adjustment of impedance parameter adjustment strategies still require significant workload and a long timeframe. Summary of the Invention

[0003] The main objective of this application is to provide a method, apparatus, medium, and device for transferring teleoperation skills by combining strategy adjustment, which aims to achieve skill transfer from human to robot by collecting and analyzing multimodal information from both master and slave ends and obtaining operational skills including strategy adjustment.

[0004] To achieve the above objectives, the first aspect of this application provides a teleoperation skill transfer method incorporating strategy adjustment, comprising: completing an operation task based on a pre-built force feedback teleoperation system to collect demonstration data, wherein the demonstration data includes interaction force data between the operator and the control device, electromyographic information of the operator, the current running trajectory of the robot, and contact force data between the robot and the environment; determining the main motion direction based on the interaction force data and the contact force data, and dividing the operation task into various stages based on the main motion direction to obtain corresponding demonstration data for each task stage; determining corresponding strategy adjustment parameters based on the demonstration data of each task stage; encoding the robot's running trajectory in stages using a dynamic motion primitive model, and calculating the mapping relationship between the strategy adjustment parameters and the running trajectory in the main motion direction using an interpolation algorithm; and planning the robot's target action based on a pre-built motion planner according to the acquired mapping relationship at the current moment, the desired position, and the contact force data, so that the robot can autonomously complete the contact operation, wherein the desired position is determined based on the current running trajectory.

[0005] Optionally, the force feedback teleoperation system includes a task end and an operation end. The task end has a robot, and the operation end has a hand controller. Before the operation task is completed based on the pre-built force feedback teleoperation system to collect demonstration data, the teleoperation skill transfer method combined with strategy adjustment further includes: based on the pre-built force feedback teleoperation system, acquiring the end-effector pose information and contact force information of the robot when completing the operation task; sending the end-effector pose information and contact force information to the hand controller; processing the end-effector pose information and contact force information to obtain the hand controller end-effector pose information and hand controller interaction force information; fixedly collecting the hand controller end-effector pose information and hand controller position change information and sending it to the robot; processing the hand controller end-effector pose information and hand controller position change information to obtain the position command of the robot controller; acquiring the robot's contact force information, transmitting the contact force information to the hand controller, and processing the contact force information to obtain the force command of the hand controller.

[0006] Optionally, determining the main motion direction based on the interaction force data and the contact force data, and dividing the task into stages based on the main motion direction, includes: performing unitized processing on the operation task based on the preprocessed interaction force data, the current running trajectory, the contact force data, and the electromyographic information to obtain corresponding unitized data; processing the interaction force data and the contact force data using a moving average algorithm to obtain preprocessed interaction force data and contact force data; calculating the difference between the preprocessed interaction force data and the contact force data, determining the main motion direction matrix based on the difference and the unit step matrix, and determining the main motion direction based on the main motion direction matrix; and dividing the operation task into multiple task stages based on the changes in the main motion direction matrix in the demonstration data.

[0007] Optionally, determining the corresponding strategy adjustment parameters based on the demonstration data of each task stage includes: determining the strategy adjustment parameters for each task stage based on the index of the electromyographic information.

[0008] Optionally, the expression of the dynamic motion primitive model, which includes nonlinear terms, involves using the dynamic motion primitive model to encode the robot's trajectory in stages and calculating the mapping relationship between the policy adjustment parameters in the main motion direction and the trajectory using an interpolation algorithm. This includes: determining the nonlinear terms for each task stage based on basis functions, a regularized system expression, and the weights of the basis functions; determining the dynamic motion primitive model for each task stage based on each nonlinear term; and calculating the mapping relationship between the policy adjustment parameters in the main motion direction and the trajectory using an interpolation algorithm based on the robot's trajectory and policy adjustment parameters for each task stage.

[0009] Optionally, the pre-built motion planner plans the robot's target action based on the acquired mapping relationship at the current moment, the desired position, and the contact force data, including: obtaining the trajectory components of the robot's current running trajectory in each direction based on the dynamic motion primitive model, and determining the robot's desired running trajectory based on the trajectory components in each direction; determining the end-effector position deviation at the current moment based on the difference between the desired running trajectory and the current running trajectory; determining the robot's desired running trajectory at the next moment based on the end-effector position deviation and the contact force data; calculating the step size at the next moment based on a preset step size calculation formula and the mapping relationship; determining the robot's end-effector position command based on the step size at the next moment, the current running trajectory, and the desired running trajectory, and obtaining the robot's planned target action based on the end-effector position command.

[0010] Optionally, the expression for the end position command is:

[0011]

[0012] Among them, s t+1 P represents the step size at the next moment. s,t+1 P represents the desired trajectory. r,t This indicates the current running trajectory.

[0013] Furthermore, to achieve the above objectives, a second aspect of this application also provides a teleoperation skill transfer device incorporating strategy adjustment, comprising: a data collection module, used to collect demonstration data by completing an operation task based on a pre-built force feedback teleoperation system, wherein the demonstration data includes interaction force data between the operator and the control device, electromyographic information of the operator, the current running trajectory of the robot, and contact force data between the robot and the environment; a task division module, used to determine the main motion direction based on the interaction force data and the contact force data, and to divide the operation task into various stages based on the main motion direction, thereby obtaining demonstration data corresponding to each task stage; a parameter determination module, used to determine corresponding strategy adjustment parameters based on the demonstration data of each task stage; a mapping module, used to encode the robot's running trajectory in stages using a dynamic motion primitive model, and to calculate the mapping relationship between the strategy adjustment parameters in the main motion direction and the running trajectory using an interpolation algorithm; and a motion planning module, used to plan the target motion of the robot based on a pre-built motion planner according to the acquired mapping relationship at the current moment, the desired position, and the contact force data, so that the robot can autonomously complete the contact operation, wherein the desired position is determined based on the current running trajectory.

[0014] To achieve the above objectives, a third aspect of this application also provides a computer-readable storage medium including instructions that, when executed on a computer, cause the computer to perform the teleoperation skill transfer method combined with strategy adjustment provided in the first aspect.

[0015] To achieve the above objectives, a fourth aspect of this application also provides an electronic device, characterized in that the electronic device includes: at least one processor, a memory, and an input / output unit; wherein the memory is used to store a computer program, and the processor is used to invoke the computer program stored in the memory to execute the teleoperation skill transfer method with combined strategy adjustment provided in the first aspect.

[0016] This application proposes a method, apparatus, medium, and device for teleoperation skill transfer that incorporates strategy adjustment. It collects demonstration data by completing an operation task based on a pre-built force feedback teleoperation system. The demonstration data includes interaction force data between the operator and the control device, the operator's electromyography (EMG) information, the robot's current trajectory, and contact force data between the robot and the environment. The main motion direction is determined based on the interaction force data and contact force data, and the task is divided into stages based on this main motion direction, resulting in the main motion direction corresponding to each stage. Strategy adjustment parameters are determined for each task stage based on the demonstration data. A dynamic motion primitive model is used to encode the mapping relationship between the strategy adjustment parameters in the main motion direction and the current trajectory in stages. Based on a pre-built motion planner, the robot's target action is planned according to the acquired mapping relationship, desired position, and contact force data at the current moment, enabling the robot to autonomously complete the contact task. This application incorporates the motion planning strategies used by humans in contact interaction tasks into human-machine skill transfer, allowing the robot to learn human strategy adjustment capabilities in tasks. This helps the robot adopt appropriate strategies during interactions with force fields or obstacles, achieving limitation and balance of trajectory deviation and contact force. Attached Figure Description

[0017] Figure 1 A flowchart is provided for an embodiment of the teleoperation skill transfer method incorporating strategy adjustments in this application;

[0018] Figure 2 This is a structural block diagram of an embodiment of the teleoperation skill transfer device incorporating strategy adjustment according to this application.

[0019] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0020] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0021] In existing technologies, human-machine skill transfer refers to the process where a human transmits their skills to a robot after providing a generalized description, thereby enabling robot programming. This can replace traditional robot programming methods and is also known as Programming by Demonstration (PbD) or Learning from Demonstration (LfD). Dynamic Motion Primitives (DMPs), as a widely used LfD method, possess many beneficial characteristics, such as good versatility and high generalization ability.

[0022] This application combines demonstration learning with a variable stiffness strategy that mimics human physical interaction behavior. The robot employing the variable stiffness strategy is influenced by the operator during physical interaction. By collecting data during teleoperation, a skill transfer algorithm can be used to acquire motor skills and corresponding changes in motion planning strategies. In multi-contact task scenarios, appropriate adversarial or compliant movements can be obtained through skill reproduction.

[0023] The teleoperation skill transfer method with strategy adjustment proposed in this application will be described in detail below with reference to embodiments and accompanying drawings.

[0024] Reference Figure 1 The teleoperation skill transfer method combined with strategy adjustment provided in the first embodiment of this application can be executed by a processor on the robot. The teleoperation skill transfer method combined with strategy adjustment may include:

[0025] S10. Based on the pre-built force feedback teleoperation system, complete the operation task to collect demonstration data, including the interaction force data between the operator and the control device, the operator's electromyographic information, the robot's current running trajectory, and the contact force data between the robot and the environment.

[0026] In this step, the processor uses a force feedback teleoperation system to collect demonstration data. Wearing electromyography (EMG) sensors, the operator controls the robot to interact with an environment of unknown mechanical properties and complete tasks. During the task, the system continuously records relevant data, including the interaction forces between the operator and the control device, the operator's EMG information, the robot's trajectory, and the contact forces between the robot and the environment.

[0027] S20. Determine the main motion direction based on the interaction force data and contact force data, and divide the operation task into various stages based on the main motion direction to obtain corresponding demonstration data for each task stage.

[0028] In this step, the processor performs trajectory preprocessing and divides the task into stages. The processor smooths the collected demonstration data and performs corresponding filtering and noise reduction on the electromyography (EMG) sensor data. The main motion direction is obtained based on the interaction force data between the operator and the control device and the contact force data between the robot and the environment, and the task is divided into stages based on changes in the main motion direction.

[0029] S30. Determine the corresponding strategy adjustment parameters based on the demonstration data of each task stage;

[0030] In this step, the processor calculates policy adjustment parameters from the demonstration data. Based on the analysis of the operator's action strategy at each moment in the demonstration data, the policy adjustment parameters for each task phase are calculated.

[0031] S40. The robot's running trajectory is encoded in stages using a dynamic motion primitive model, and the mapping relationship between the strategy adjustment parameters in the main motion direction and the running trajectory is calculated using an interpolation algorithm.

[0032] In this step, the processor encodes the trajectory and strategy adjustment parameters in stages. For each task stage, the processor uses a dynamic motion primitive model to encode the relationship between the desired trajectory of the robot's end effector and the current trajectory, and records the relationship between the strategy adjustment parameters and the position of the main motion direction.

[0033] S50. Based on the pre-built motion planner, the robot plans the target action according to the obtained current mapping relationship, expected position and contact force data, so that the robot can autonomously complete the contact operation. The expected position is determined based on the current running trajectory.

[0034] In this step, the processor completes the task through skill reproduction. First, a motion planner for the robot is constructed. During the task, the robot obtains the desired trajectory and policy adjustment parameters corresponding to the current position through the dynamic primitive model in step S40. The motion planner then plans the target action according to the desired position, the current position, and the environmental contact force, following the current policy, enabling the robot to autonomously complete the contact operation.

[0035] In summary, this application proposes a teleoperation skill transfer method that combines strategy adjustment. It collects demonstration data by completing an operation task based on a pre-built force feedback teleoperation system. This demonstration data includes interaction force data between the operator and the control device, the operator's electromyography (EMG) information, the robot's current trajectory, and contact force data between the robot and the environment. The main motion direction is determined based on the interaction and contact force data, and the task is divided into stages based on this main motion direction, yielding the main motion direction corresponding to each stage. Strategy adjustment parameters are determined for each task stage based on the demonstration data. A dynamic motion primitive model is used to encode the mapping relationship between the strategy adjustment parameters in the main motion direction and the current trajectory in stages. Based on a pre-built motion planner, the robot's target action is planned according to the acquired mapping relationship, desired position, and contact force data at the current moment, enabling the robot to autonomously complete the contact operation. The desired position is determined based on the current trajectory. This application incorporates the motion planning strategies used by humans in contact interaction tasks into human-machine skill transfer, enabling the robot to learn human strategy adjustment capabilities during tasks. This helps the robot adopt appropriate strategies during interactions with force fields or obstacles, achieving limitation and balance of trajectory deviation and contact force. Based on the modeling and parameterization of the interaction strategy, this application employs dynamic motion primitives to learn motion trajectories and strategy changes, enabling the learning and reproduction of human operational skills and improving the completeness of these skills. In the trajectory data preprocessing stage, this application divides the task into stages and performs segmented skill learning based on these stages. Furthermore, the motion planning also references the main motion direction corresponding to each stage, making the system more adaptable to different environments and task objectives.

[0036] In embodiments of this application, the force feedback teleoperation system includes a task end and an operation end. The task end has a robot, and the operation end has a hand controller. Before completing the operation task based on the pre-built force feedback teleoperation system to collect demonstration data, the teleoperation skill transfer method combined with strategy adjustment further includes:

[0037] Based on a pre-built force feedback telesystem, the robot acquires end-effector pose and contact force information when completing the operation task.

[0038] Send the end-effector pose information and contact force information to the hand controller;

[0039] Process the end-effector pose information and contact force information to obtain the end-effector pose information and hand controller interaction force information;

[0040] The system collects the end-effector pose information and position change information of the hand controller and sends them to the robot.

[0041] Process the end-effector pose information and the change in the end-effector position to obtain the position command of the robot controller;

[0042] The robot acquires contact force information, transmits the contact force information to the hand controller, processes the contact force information, and obtains the force command from the hand controller.

[0043] Specifically, the specific implementation of the present invention will be described with a specific task example. The robot in the example is UR5, the control device used by the human operator is the Force Dimension omega.6 hand controller, and the collection of electromyographic signals is accomplished by an electromyographic wristband. The task objective is to move the robot's end effector from one point in space to another on a plane. In this process, it is necessary to conform to the surface of hard obstacles and overcome soft obstacles. The present invention will now be further described with reference to the accompanying drawings:

[0044] The processor has implemented the force feedback teleoperation system through the above steps. Furthermore, the construction process of the force feedback teleoperation system may include the following specific execution steps:

[0045] Construct a robot with a task end and an operation end, and deploy a force sensor at the end of the robot's end. The robot's end-effector pose information is represented by a homogeneous matrix T, and the contact force is acquired by the force sensor, which is represented by F. r The expression is shown in equation (1):

[0046]

[0047] In equation (1), R r Let P be the rotation matrix corresponding to the robot's end effector posture. r =[x r y r z r ] T F represents the position coordinates of the robot's end effector. r A force component comprising three directions, denoted as F. r =[F xr F yr F zr ] T The processor sends the end-effector pose information and contact force information to the hand controller at the operating end.

[0048] The hand controller will be set to force feedback mode. Without a force command, the operator can freely translate and rotate the hand controller by holding its end. Force commands can be set to provide force feedback at the end of the hand controller. The processor can read the current hand controller end pose information Q and the interaction force F between the operator and the hand controller end. q As shown below:

[0049]

[0050] In equation (2), R q P is the rotation matrix corresponding to the hand controller's posture. q =[x q y q z q ] T F represents the position coordinates of the end of the hand controller. q A force component comprising three directions, denoted as F. q =[F xq F yq F zq ] T .

[0051] During the initialization of the teleoperation system, the robot end effector and the hand controller end effector will be placed in the same pose R0, and their initial positions P will be recorded upon completion. q0 The hand controller acquires end-effector pose data at a fixed frequency and records the change in position at time t relative to the initial position, ΔP. t The change in position ΔP of the hand controller is transmitted to the robot. t As shown in equation (3)::(3)

[0052] The task-side robot calculates the corresponding command values ​​based on the received attitude and position changes and inputs them into the robot controller and the force command from the hand controller. for:

[0053]

[0054] The task-side robot collects contact force information F with the environment during its work. r,t It is then transmitted to the operating terminal as a force command. Input the hand controller, the hand controller's force command The expression is shown in the following formula:

[0055]

[0056] In step S10, the processor collects demonstration data using a force feedback teleoperation system. Wearing electromyography (EMG) sensors, the operator controls the robot to interact with an environment of unknown mechanical properties and complete tasks using the pre-built force feedback teleoperation system. During the task, the force feedback teleoperation system continuously records relevant data, including the interaction force F between the operator and the control device. h,t Operator's electromyographic information A t The robot's trajectory P r,t The contact force F between the robot and the environment r,t Thus, the original demonstration dataset L is obtained, as shown in equation (6):

[0057] L = {Pr,t A t ,F h,t ,F r,t |t=0,1,2……t end} (5)

[0058] In the formula, the subscript t represents the data corresponding to the data, and the total number of data points is represented by t. end express.

[0059] The processor collects demonstration data by executing step S10, thereby acquiring motor skills and corresponding action planning strategy changes using the skill transfer algorithm, and obtaining appropriate adversarial or compliant movements from skill reproduction in multi-contact task scenarios.

[0060] In embodiments of this application, step S20 may include the following execution process:

[0061] S201. Based on the preprocessed interaction force data, current running trajectory, contact force data and electromyographic information, the operation task is processed into units to obtain the corresponding unit data;

[0062] S202. The moving average algorithm is used to process the interaction force data and contact force data to obtain preprocessed interaction force data and contact force data.

[0063] S203. Calculate the difference between the preprocessed interaction force data and the contact force data, determine the main motion direction matrix based on the difference and the unit step matrix, and determine the main motion direction based on the main motion direction matrix;

[0064] S204. Based on the changes in the main motion direction matrix in the demonstration data, the operation task is divided into multiple task stages.

[0065] The processor performs trajectory preprocessing and task segmentation through the above steps S201-S204. In the trajectory data preprocessing stage, the application divides the task into stages and performs segmented skill learning based on the divided stages. The main motion direction corresponding to the stage is also referenced in the motion planning, so that the system has better adaptability to different environments and task objectives.

[0066] Specifically, the steps of the processor performing trajectory preprocessing and dividing the task into stages may include the following execution process:

[0067] Trajectory preprocessing and task phase division are performed. Data marked with an asterisk (*) is used to indicate processed data. The trajectory is smoothed using polynomial interpolation and velocity planning preprocessing to obtain a continuous and smooth demonstration trajectory. The contact force and interaction force in the demonstration data were preprocessed using a moving average. The electromyography signal intensities of each channel were weighted, summed, and then processed using a moving average to obtain the result. The result is L * This is represented as shown in equation (7):

[0068]

[0069] The main motion direction is obtained by the difference between the interaction force data between the operator and the control device and the contact force data between the robot and the environment. The calculation process is shown in equations (8) and (9):

[0070]

[0071] In equation (7), F d,t =[F xd,t F yd,t F zd,t ] T Let D be the difference between the interaction force data between the operator and the control device at time t and the contact force data between the robot and the environment. In equation (8), D t Let H(x) represent the main motion direction at time t, and let H(x) be the unit step function, defined as shown in equation (10):

[0072]

[0073] According to D t The changes in the value throughout the demonstration data can divide the task into different stages, and obtain the main motion direction corresponding to each stage, which is represented by equation (11):

[0074]

[0075] In embodiments of this application, step S30 may include the following execution process:

[0076] The strategy adjustment parameters for each task stage are determined based on the index of electromyographic information.

[0077] In this step, the processor can calculate the strategy adjustment parameters from the demonstration data.

[0078] Specifically, the processor can calculate the strategy adjustment parameters for each task stage based on the processed electromyographic information in the demonstration data, as shown in Equation (12):

[0079]

[0080] In equation (11), K t Let t be the strategy adjustment parameter at time t, where k and γ are parameters set during the initialization phase based on data characteristics and empirical knowledge.

[0081] In embodiments of this application, step S40 may include the following execution process:

[0082] S401. Determine the nonlinear terms for each task stage based on basis functions, regular system expressions, and the weights of basis functions;

[0083] S402. Determine the dynamic motion primitive model for each task stage based on the corresponding nonlinear terms;

[0084] S403. Based on the robot's running trajectory and strategy adjustment parameters at each task stage, calculate the mapping relationship between the strategy adjustment parameters and the running trajectory in the main motion direction using an interpolation algorithm.

[0085] The processor encodes the trajectory and strategy adjustment parameters in stages through the steps S401-S403 described above.

[0086] Specifically, the steps by which the processor encodes the trajectory and policy adjustment parameters in stages may include the following execution process:

[0087] For each task stage, the relationship between the robot's end effector's desired trajectory and time is encoded using dynamic motion primitives. The form of the dynamic motion primitives is shown in Equation (13):

[0088]

[0089] Where p represents each component of the position information P, that is, for the x, y, z dynamic motion primitives, the same form is satisfied. Replacing p yields the specific form of each component, where... g is the set target value, τ is the set time scaling constant, and α p and β p Let f be a constant and f be a nonlinear term, constructed as shown in equation (14).

[0090]

[0091] In the formula, θ comes from the regular system, and θ = 1 corresponds to the initial state. The expression of the regular system is shown in equation (15):

[0092]

[0093] In equation (13), p0 represents the initial state, ω i For different basis functions Ψ i The weights of (θ), where N is the number of basis functions, and the basis functions Ψ i The expression for (θ) is shown in (16).

[0094]

[0095] In equation (15), ci The center position of the basis functions needs to be obtained through learning. After constructing the dynamic motion primitives, the data obtained in step 4 needs to be used for learning to obtain the various parameters in the model.

[0096] Extract the results obtained in step 3 Let t1 represent The total duration of this data segment is first determined by setting α based on experience. p ,α θ ,β p N. For learning f, a locally weighted regression method is used. The required learning value of f can be obtained through transformation equation (12). t As shown in equation (17):

[0097]

[0098] In equation (16) p d For P r The components of the basis functions. Given that the basis functions are uniformly distributed over the entire time interval, and defined as ξ(t) = θ(t)(g-p0), the weights ω of the basis functions can be obtained. i As shown in equation (18):

[0099]

[0100] in:

[0101]

[0102] Based on the main motion direction, trajectory, and strategy adjustment parameters in each stage, the mapping relationship between the strategy adjustment parameters and the trajectory is obtained by interpolation, as shown in equation (20):

[0103]

[0104] In equation (19), through Obtain the position coordinates of the trajectory in the active driving direction at each moment during each task phase, h n () indicates the mapping relationship between the strategy adjustment parameters and the running trajectory in the nth stage of the corresponding task.

[0105] In embodiments of this application, step S50 may include the following execution process:

[0106] S501. Based on the dynamic motion primitive model, obtain the trajectory components of the robot's current running trajectory in each direction at the current moment, and determine the robot's expected running trajectory based on the trajectory components in each direction.

[0107] S502. Determine the end position deviation and strategy adjustment parameters at the current moment based on the difference between the expected running trajectory and the current running trajectory;

[0108] S503. Determine the robot's expected trajectory at the next moment based on strategy adjustment parameters, end-effector position deviation, and contact force data;

[0109] S504. Calculate the step size for the next time step based on the preset step size calculation formula and mapping relationship;

[0110] S505. Based on the step size of the next moment, the current running trajectory, and the expected running trajectory, determine the robot's end-effector position command, and obtain the robot's planned target action based on the end-effector position command.

[0111] The processor completes the task through skill reproduction through the above steps S501-S505. This application uses dynamic motion primitives to learn motion trajectory and strategy changes, which can realize the learning and reproduction of human operation skills and improve the integrity of operation skills.

[0112] Specifically, the steps to complete a task through skill reproduction can include the following execution process:

[0113] Skills are reproduced sequentially according to the task order to complete the task. Within each stage, the expected trajectory at that moment is first obtained using the dynamic primitive model from step 5; the target position g is determined based on the task information for that stage. r And the task's time scaling parameter τ. This is determined by α. p ,β p ,ω i ,Ψ i ,α θ Combining formula (21), the components p of the robot end effector reference trajectory in each direction can be calculated:

[0114]

[0115] The desired trajectory P of the robot is obtained from equation (20). d,t =[P dx,t P dy,t P dz,t ] T And calculate the current end position deviation E. t The strategy adjustment parameters corresponding to the current position are shown in equations (22) and (23), respectively:

[0116] E t =P d,t -P r,t (twenty one)

[0117]

[0118] The expected trajectory P of the robot's end effector at the next moment s,t+1 The result is calculated by combining the position deviation and contact force of the robot end effector with the strategy adjustment parameters, as shown in equation (24):

[0119]

[0120] In equation (23), F r,t K0 is a fixed parameter representing the contact force measured by the robot's end effector force sensor in each direction at time t.

[0121] In P s,t+1 ≤P s,min At that time, step size s t+1 =0; in P s,t+1 >P s,min At that time, step size s t+1 Calculate according to formula (25):

[0122]

[0123] In equation (24), s m and s l These are all parameters pre-set according to the task, namely the basic compensation value and the step size limit value. Based on the step size s... t+1 and the desired direction of motion P s,t+1 It can calculate the robot's end-effector position command. The robot is controlled as shown in the equation.

[0124]

[0125] Among them, s t+1 P represents the step size at the next moment. s,t+1 P represents the desired trajectory. r,t This indicates the current running trajectory.

[0126] refer to Figure 2 Based on the above embodiments, this application also provides a teleoperation skill transfer device incorporating strategy adjustment. The teleoperation skill transfer device 100 incorporating strategy adjustment includes a data collection module 101, a task division module 102, a parameter determination module 103, a mapping module 104, and a motion planning module 105, wherein...

[0127] The data collection module 101 is used to complete the operation task based on the pre-built force feedback teleoperation system to collect demonstration data. The demonstration data includes the interaction force data between the operator and the control device, the operator's electromyographic information, the robot's current running trajectory, and the contact force data between the robot and the environment.

[0128] The task division module 102 is used to determine the main motion direction based on the interaction force data and contact force data, and to divide the operation task into various stages based on the main motion direction, so as to obtain corresponding demonstration data for each task stage.

[0129] The parameter determination module 103 is used to determine the corresponding strategy adjustment parameters based on the demonstration data of each task stage;

[0130] The mapping module 104 is used to encode the robot's running trajectory in stages using a dynamic motion primitive model, and to calculate the mapping relationship between the strategy adjustment parameters in the main motion direction and the running trajectory through an interpolation algorithm;

[0131] The motion planning module 105 is used to plan the robot's target motion based on the pre-built motion planner and the obtained current mapping relationship, desired position and contact force data, so that the robot can autonomously complete the contact operation. The desired position is determined based on the current running trajectory.

[0132] Based on the above embodiments, this application also provides a computer-readable storage medium including instructions that, when executed on a computer, cause the computer to implement the teleoperation skill transfer method with combined strategy adjustment provided by any of the aforementioned methods.

[0133] Based on the above embodiments, this application also provides an electronic device, which includes: at least one processor, a memory, and an input / output unit; wherein the memory is used to store a computer program, and the processor is used to call the computer program stored in the memory to execute the teleoperation skill transfer method with combined strategy adjustment provided by any of the foregoing methods.

[0134] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for teleoperation skill transfer incorporating strategy adjustments, characterized in that, include: The operation task is completed based on a pre-built force feedback teleoperation system to collect demonstration data, which includes the interaction force data between the operator and the control device, the operator's electromyography information, the robot's current running trajectory, and the contact force information between the robot and the environment. The main motion direction is determined based on the interaction force data and the contact force information, and the operation task is divided into various stages based on the main motion direction, so that each task stage has corresponding demonstration data. Determine the corresponding strategy adjustment parameters based on the demonstration data of each task phase; The robot's trajectory is encoded in stages using a dynamic motion primitive model, and the mapping relationship between the strategy adjustment parameters in the main motion direction and the trajectory is calculated using an interpolation algorithm. Based on the pre-built motion planner, the robot plans the target action according to the obtained current mapping relationship, desired position and contact force information, so that the robot can autonomously complete the contact operation. The desired position is determined based on the current running trajectory. The step of determining the main motion direction based on the interaction force data and the contact force information, and dividing the task into stages based on the main motion direction, includes: Based on the preprocessed interaction force data, the current running trajectory, the contact force information, and the electromyographic information, the operation task is processed into units to obtain corresponding unitized data; The interaction force data and the contact force information are processed using a moving average algorithm to obtain preprocessed interaction force data and contact force information; Calculate the difference between the preprocessed interaction force data and the contact force information, determine the main motion direction matrix based on the difference and the unit step matrix, and determine the main motion direction based on the main motion direction matrix; Based on the changes in the main motion direction matrix in the demonstration data, the operation task is divided into multiple task stages; The determination of corresponding strategy adjustment parameters based on demonstration data from each task phase includes: The strategy adjustment parameters for each task stage are determined based on the index of the electromyographic information.

2. The teleoperation skill transfer method combined with strategy adjustment as described in claim 1, characterized in that, The force feedback teleoperation system includes a task terminal and an operation terminal. The task terminal has a robot, and the operation terminal has a hand controller. Operational tasks are performed within the pre-built force feedback teleoperation system to collect demonstration data. The teleoperation skill transfer method incorporating strategy adjustments further includes: Based on the pre-constructed force feedback teleoperation system, the end-effector pose information and contact force information of the robot when completing the operation task are obtained; The end-effector pose information and the contact force information are sent to the hand controller; Process the end-effector pose information and the contact force information to obtain the end-effector pose information and the hand controller interaction force information; The system collects the end-effector pose information and the change in position of the hand controller and sends them to the robot. Process the end-effector pose information and the change in the end-effector position to obtain the position command of the robot controller; The contact force information of the robot is acquired, the contact force information is transmitted to the hand controller, and the contact force information is processed to obtain the force command of the hand controller.

3. The teleoperation skill transfer method combined with strategy adjustment as described in claim 1, characterized in that, The expression of the dynamic motion primitive model includes nonlinear terms. The process of using the dynamic motion primitive model to encode the robot's trajectory in stages and calculating the mapping relationship between the strategy adjustment parameters in the main motion direction and the trajectory using an interpolation algorithm includes: The nonlinear terms for each task stage are determined based on basis functions, regular system expressions, and the weights of the basis functions. The dynamic motion primitive model for each task stage is determined based on the corresponding nonlinear terms. Based on the robot's running trajectory and strategy adjustment parameters at each task stage, the mapping relationship between the strategy adjustment parameters and the running trajectory in the main motion direction is calculated using an interpolation algorithm.

4. The teleoperation skill transfer method combined with strategy adjustment as described in claim 1, characterized in that, The pre-built motion planner plans the robot's target action based on the acquired mapping relationship at the current moment, the desired position, and the contact force information, including: Based on the dynamic motion primitive model, the trajectory components of the robot's current running trajectory in each direction are obtained at the current moment, and the expected running trajectory of the robot is determined based on the trajectory components in each direction. The terminal position deviation and strategy adjustment parameters at the current moment are determined based on the difference between the expected running trajectory and the current running trajectory. The robot's expected trajectory at the next moment is determined based on the strategy adjustment parameters, the end-effector position deviation, and the contact force information. The step size for the next moment is calculated based on the preset step size calculation formula and the mapping relationship. Based on the step size at the next moment, the current running trajectory, and the expected running trajectory, the robot's end-effector position command is determined, and the robot's planned target action is obtained based on the end-effector position command.

5. The teleoperation skill transfer method combined with strategy adjustment as described in claim 4, characterized in that, The expression for the end position command is: in, Indicates the step size at the next moment. Indicates the desired trajectory. This indicates the current running trajectory.

6. A teleoperation skill transfer device incorporating strategy adjustment, characterized in that, The teleoperation skill transfer method combining strategy adjustment as described in any one of claims 1-5 includes: The data collection module is used to complete operation tasks based on the pre-built force feedback teleoperation system to collect demonstration data. The demonstration data includes the interaction force data between the operator and the control device, the operator's electromyography information, the robot's current running trajectory, and the contact force information between the robot and the environment. The task division module is used to determine the main motion direction based on the interaction force data and the contact force information, and to divide the operation task into various stages based on the main motion direction, so as to obtain corresponding demonstration data for each task stage. The parameter determination module is used to determine the corresponding strategy adjustment parameters based on the demonstration data of each task stage; The mapping module is used to encode the robot's running trajectory in stages using a dynamic motion primitive model, and to calculate the mapping relationship between the strategy adjustment parameters in the main motion direction and the running trajectory through an interpolation algorithm. The motion planning module is used to plan the robot's target motion based on a pre-built motion planner, according to the obtained mapping relationship at the current moment, the desired position, and the contact force information, so that the robot can autonomously complete the contact operation. The desired position is determined based on the current running trajectory.

7. A computer-readable storage medium, characterized in that, It includes instructions that, when executed on a computer, cause the computer to perform the teleoperation skill transfer method with combined strategy adjustment as described in any one of claims 1 to 5.

8. An electronic device, characterized in that, The electronic device includes: At least one processor, memory, and input / output unit; The memory is used to store computer programs, and the processor is used to invoke the computer programs stored in the memory to execute the teleoperation skill transfer method with combined strategy adjustment according to any one of claims 1 to 5.

Images