Robot URDF model free combination and constraint management method and system

Abstract

The invention relates to the technical field of robots, and discloses a robot URDF model free combination and constraint management method and system, and the method comprises the steps: obtaining user demand data, and determining a selectable part set of a to-be-assembled robot according to the user demand data; calculating the matching degree of selectable components at each assembly position in the selectable component set, and establishing a robot assembly scheme according to the matching degree of the selectable components at each assembly position; predicting a prediction task execution effect of each robot assembly scheme, and screening out an optimal assembly scheme according to the prediction task execution effect; and performing task simulation on the optimal assembly scheme based on a URDF model, establishing a trajectory constraint set of the robot according to a task simulation result, and controlling the robot to execute the task according to the trajectory constraint set. The trajectory constraint set integrates kinematics constraint and perception constraint, integrated free combination and cooperative constraint management of a robot actuator and a sensor are achieved, and the robot is more reasonably and reliably controlled to complete complex tasks.

Description

Technical Field

[0001] This application relates to the field of robotics technology, and more specifically, to a method and system for free combination and constraint management of robot URDF models. Background Technology

[0002] In traditional applications such as industry, robots face relatively fixed tasks and environments, often employing mobile robots and articulated robots with relatively fixed configurations. These robots can complete specific tasks within specific work scenarios and with known structural information. However, emerging fields typically involve diverse and unstructured tasks. Therefore, robots need to be fully autonomously constructed with the ability to freely combine multiple modules to meet different needs.

[0003] Current fixed-structure robots are limited by their single form, making it difficult to meet the diverse and modular task requirements. Summary of the Invention

[0004] This invention provides a method and system for the free combination and constraint management of URDF models of robots, to solve the problem that fixed-structure robots in the prior art are difficult to meet the diverse and multi-module free combination task requirements, including: The process involves: acquiring user demand data; determining a set of selectable components for the robot to be assembled, including an actuator set and a sensor set; calculating the matching degree of selectable components at each assembly position in the selectable component set; establishing a robot assembly scheme based on the matching degree of selectable components at each assembly position; predicting the predicted task execution effect of each robot assembly scheme; selecting the optimal assembly scheme based on the predicted task execution effect; performing task simulation on the optimal assembly scheme based on the URDF model; establishing a trajectory constraint set for the robot based on the task simulation results; and controlling the robot to execute tasks based on the trajectory constraint set.

[0005] Furthermore, the step of determining the set of selectable parts for the robot to be assembled based on user demand data includes: establishing a natural language model; interpreting the user demand data based on the natural language model; determining the functional requirement text of the parts corresponding to each position based on the interpretation results; interpreting the functional description text of the parts in the current robot part library based on the natural language model to obtain the component functional description text of each part; matching the functional requirement text of each part with the functional description text of each part to obtain the selectable parts at each position; and establishing a set of selectable parts based on the selectable parts at each position.

[0006] Further, the step of matching the functional requirement text of each component with the functional description text of each component to obtain selectable components at each position includes: vectorizing the functional requirement text and the functional description text of each component to obtain a component functional requirement vector and a component functional description vector; calculating the cosine similarity between the component functional requirement vector and the component functional description vector, and filtering out component functional description vectors with a cosine similarity greater than a first preset threshold; and using the component corresponding to the filtered component functional description vector as the selectable component at the position corresponding to the component functional requirement vector to obtain selectable components at all positions.

[0007] Furthermore, the step of establishing a robot assembly scheme based on the matching degree of selectable components at each assembly position includes: obtaining the component assembly point of any target selectable component; matching the component assembly point of the target selectable component with the component assembly points of all selectable components at the corresponding assembly position; obtaining the number of matching assembly points of matching components at the assembly position that can match the target selectable component; counting the assembly point type to which the matching assembly point of each matching component belongs; calculating the matching degree between the target selectable component and the matching component based on the assembly point type of the target selectable component and the matching component and the number of matching assembly points of each assembly point type; filtering out matching components with a matching degree greater than a second preset threshold; and establishing a robot assembly scheme based on the filtered matching components.

[0008] Further, the step of calculating the matching degree between the target selectable component and the matching component based on the assembly point type of the target selectable component and the matching component and the number of matching assembly points for each assembly point type includes: determining the matching degree between the target selectable component and the matching component according to the matching degree formula, wherein the matching degree formula is, , in, The matching degree between the target component and the matching component can be selected. For the first Importance weight of assembly point type , , For the target, selectable components for the first The preset standard quantity requirement for assembly point type For the first matching component Number of matching assembly points of the assembly point type.

[0009] Furthermore, the step of predicting the performance of each robot assembly scheme and selecting the optimal assembly scheme based on the predicted performance includes: establishing a robot performance prediction model, the input of which is the component parameters of each component after assembly, and the output is the predicted performance; constructing a training set based on historical robot task execution data, establishing a loss function, and training the robot performance prediction model based on the training set and the loss function; inputting the component parameters of the current assembly scheme into the trained robot performance prediction model to obtain the predicted performance of each assembly scheme.

[0010] Furthermore, the robot performance prediction model includes a component feature input module, a Transformer encoder module, and an MLP module. The Transformer encoder module is used to capture the complex dependencies between the parameters of each component and output the feature representation after the component parameters are fused. The MLP module is used to map the fused feature representation to the final task execution effect.

[0011] Furthermore, the loss function is: , in, For loss function, To predict the success rate of the task, For actual task success rate, This refers to the actual time it took to complete the task. To predict task completion time, , To preset the first weight and the second weight, .

[0012] Furthermore, the step of performing task simulation on the optimal assembly scheme based on the URDF model and establishing a robot trajectory constraint set based on the task simulation results includes: determining robot task data based on user requirement data and determining task constraint boundaries based on the robot task data; establishing a URDF model of the optimal assembly scheme and importing the URDF model of the optimal assembly scheme into a physical simulation environment for task simulation; determining the robot trajectory when performing the task based on the robot task simulation results; setting robot trajectory constraints when the robot trajectory reaches the constraint boundary during task execution; and establishing a trajectory constraint set based on all robot trajectory constraints.

[0013] To achieve the above objectives, the present invention also provides a robot URDF model free combination and constraint management system, comprising: The system comprises the following modules: an acquisition module for acquiring user demand data and determining a set of selectable components for the robot to be assembled, including a set of actuators and a set of sensors; a matching module for calculating the matching degree of selectable components at each assembly position in the selectable component set and establishing a robot assembly scheme based on the matching degree of selectable components at each assembly position; an assembly module for predicting the predicted task execution effect of each robot assembly scheme and selecting the optimal assembly scheme based on the predicted task execution effect; and a simulation module for performing task simulation on the optimal assembly scheme based on the URDF model, establishing a trajectory constraint set for the robot based on the task simulation results, and controlling the robot to execute tasks based on the trajectory constraint set.

[0014] The beneficial effects of this invention are as follows: By applying the above technical solutions, this invention establishes a set of selectable robot components based on user needs, assembles the robot by matching the components at each assembly position, and realizes the free assembly of multiple robot modules. At the same time, by performing task simulation on the assembled robot, a set of trajectory constraints for the robot is established, which integrates kinematic constraints and perception constraints, realizing the integrated free combination and collaborative constraint management of robot actuators and sensors, and more rationally and reliably controlling the robot to complete complex tasks. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 The overall flowchart of a robot URDF model free combination and constraint management method proposed in an embodiment of the present invention is shown; Figure 2 The diagram shows a schematic of a robot URDF model free combination and constraint management system proposed in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0018] This application provides a method for free combination and constraint management of a robot URDF model, such as... Figure 1As shown, it includes: S101, Obtain user demand data, and determine the set of selectable parts for the robot to be assembled based on the user demand data. The set of selectable parts includes a set of actuators and a set of sensors. In some embodiments of this application, determining the set of selectable parts for the robot to be assembled based on user demand data includes: establishing a natural language model; interpreting the user demand data based on the natural language model; determining the component functional requirement text for each position corresponding to the user demand data based on the interpretation results; interpreting the functional description text of the components in the current robot component library based on the natural language model to obtain the component functional description text of each component; matching the functional requirement text of each component with the functional description text of each component to obtain the selectable parts for each position; and establishing a set of selectable parts based on the selectable parts for each position.

[0019] In this embodiment, a natural language model is established to interpret the user's input requirement data, and the functional requirement text of the robot parts corresponding to the user's requirements at each location is interpreted. The robot parts specifically include actuator parts and sensor parts. At the same time, the functional description text of all parts in the robot part library is interpreted by the natural language model. By matching the functional requirement text of each part with the functional description text of each part, the selectable parts at each location with a high degree of matching with the user's requirements are obtained, and a set of selectable parts is established.

[0020] In some embodiments of this application, the step of matching the functional requirement text of each component with the functional description text of each component to obtain selectable components at each position includes: vectorizing the functional requirement text and the functional description text of each component to obtain a component functional requirement vector and a component functional description vector; calculating the cosine similarity between the component functional requirement vector and the component functional description vector, and filtering out component functional description vectors with a cosine similarity greater than a first preset threshold; and using the component corresponding to the filtered component functional description vector as the selectable component at the position corresponding to the component functional requirement vector to obtain selectable components at all positions.

[0021] In this embodiment, by vectorizing the functional requirement text and functional description text of each component and calculating the cosine similarity, the components with high similarity to the functional requirement text of each component are extracted as selectable components.

[0022] S102 calculates the matching degree of selectable parts at each assembly position in the selectable parts set, and establishes a robot assembly scheme based on the matching degree of selectable parts at each assembly position. In some embodiments of this application, the step of establishing a robot assembly scheme based on the matching degree of selectable components at each assembly position includes: obtaining the component assembly point of any target selectable component; matching the component assembly point of the target selectable component with the component assembly points of all selectable components at the corresponding assembly position; obtaining the number of matching assembly points of matching components at the assembly position that can match the target selectable component; counting the assembly point type to which the matching assembly point of each matching component belongs; calculating the matching degree between the target selectable component and the matching component based on the assembly point type of the target selectable component and the matching component and the number of matching assembly points of each assembly point type; filtering out matching components with a matching degree greater than a second preset threshold; and establishing a robot assembly scheme based on the filtered matching components.

[0023] In this embodiment, a selectable component contains multiple assembly points. Each assembly point is divided according to the assembly type of the selectable component. The assembly point type specifically includes interface type, assembly position, and assembly method. The matching degree between the two components is calculated based on the matching assembly point type of the matching component that can match the assembly point of the target selectable component. This allows for the selection of matching components that match the target selectable component at the assembly position. Each matching component is then assembled with the target selectable component to form multiple robot assembly schemes.

[0024] In some embodiments of this application, calculating the matching degree between the target selectable component and the matching component based on the assembly point type of the target selectable component and the matching component and the number of matching assembly points for each assembly point type includes: determining the matching degree between the target selectable component and the matching component according to a matching degree formula, wherein the matching degree formula is as follows: , in, The matching degree between the target component and the matching component can be selected. For the first Importance weight of assembly point type , , For the target, selectable components for the first The preset standard quantity requirement for assembly point type For the first matching component Number of matching assembly points of the assembly point type.

[0025] In this embodiment, the matching degree between the target selectable component and the matching component is calculated by the number of matching assembly points corresponding to each assembly point type. and Shishi This makes the calculation results more accurate, enabling precise calculation of component matching degree.

[0026] S103, predict the performance of each robot assembly scheme in the prediction task, and select the optimal assembly scheme based on the performance of the prediction task. In some embodiments of this application, the step of predicting the performance of each robot assembly scheme and selecting the optimal assembly scheme based on the predicted performance includes: establishing a robot performance prediction model, the input of which is the component parameters of each component after assembly, and the output is the predicted performance; constructing a training set based on historical robot task execution data, establishing a loss function, and training the robot performance prediction model based on the training set and the loss function; inputting the component parameters of the current assembly scheme into the trained robot performance prediction model to obtain the predicted performance of each assembly scheme.

[0027] In some embodiments of this application, the robot performance prediction model includes a component feature input module, a Transformer encoder module, and an MLP module. The Transformer encoder module is used to capture the complex dependencies between the parameters of each component and output the feature representation after the component parameters are fused. The MLP module is used to map the fused feature representation to the final task execution effect.

[0028] In this embodiment, the robot's task to be performed is determined by user demand data, and a robot performance prediction model is constructed by historical robot task execution data to predict the task execution effect of each assembly scheme. The component parameters specifically include the selected component combination, linkage parameters, sensor layout, task scenario, etc. The task execution effect includes the task success rate and task completion time, which facilitates the subsequent acquisition of complete robot trajectory, torque and perception data.

[0029] In some embodiments of this application, the loss function is: , in, For loss function, To predict the success rate of the task, For actual task success rate, This refers to the actual time it took to complete the task. To predict task completion time, , To preset the first weight and the second weight, .

[0030] In this embodiment, the loss function is formed by weighted summation of the task success rate prediction loss and the task completion time prediction loss, which effectively improves the accuracy of the model prediction.

[0031] S104. Based on the URDF model, the optimal assembly scheme is simulated for task execution. The trajectory constraint set of the robot is established according to the task simulation results, and the robot is controlled to execute the task according to the trajectory constraint set.

[0032] In some embodiments of this application, the step of performing task simulation on the optimal assembly scheme based on the URDF model and establishing a robot trajectory constraint set based on the task simulation results includes: determining robot task data based on user requirement data, determining task constraint boundaries based on the robot task data; establishing a URDF model of the optimal assembly scheme, importing the URDF model of the optimal assembly scheme into a physical simulation environment for task simulation; determining the robot trajectory when performing the task based on the robot task simulation results, setting robot trajectory constraints when the robot trajectory reaches the constraint boundary, and establishing a trajectory constraint set based on all robot trajectory constraints.

[0033] In this embodiment, task constraint boundaries are set based on the tasks to be executed corresponding to user demand data to ensure that the task target is always within the field of view and to avoid sensor self-occlusion, etc. An URDF model of the optimal assembly scheme is established and imported into the physical simulation environment for detailed task simulation to obtain complete robot trajectory, torque and perception data. The robot's running trajectory is recorded and the trajectory position data when the running trajectory exceeds the constraint boundary is marked. The trajectory constraints of the robot are set based on the trajectory position data when the running trajectory exceeds the constraint boundary, thereby establishing a trajectory constraint set and realizing reasonable control of the robot.

[0034] Based on the same technological concept, such as Figure 2 As shown, the present invention also provides a robot URDF model free combination and constraint management system, including: The system comprises the following modules: an acquisition module for acquiring user demand data and determining a set of selectable components for the robot to be assembled, including a set of actuators and a set of sensors; a matching module for calculating the matching degree of selectable components at each assembly position in the selectable component set and establishing a robot assembly scheme based on the matching degree of selectable components at each assembly position; an assembly module for predicting the predicted task execution effect of each robot assembly scheme and selecting the optimal assembly scheme based on the predicted task execution effect; and a simulation module for performing task simulation on the optimal assembly scheme based on the URDF model, establishing a trajectory constraint set for the robot based on the task simulation results, and controlling the robot to execute tasks based on the trajectory constraint set.

[0035] By applying the above technical solutions, this invention acquires user demand data and determines the set of selectable parts for the robot to be assembled based on this data; calculates the matching degree of selectable parts at each assembly position in the selectable part set, and establishes a robot assembly scheme based on the matching degree of selectable parts at each assembly position; predicts the predicted task execution effect of each robot assembly scheme, and selects the optimal assembly scheme based on the predicted task execution effect; performs task simulation on the optimal assembly scheme based on the URDF model, establishes the robot's trajectory constraint set based on the task simulation results, and controls the robot to execute tasks based on the trajectory constraint set. The trajectory constraint set integrates kinematic constraints and perceptual constraints, realizing the integrated free combination and collaborative constraint management of robot actuators and sensors, and more rationally and reliably controlling the robot to complete complex tasks.

[0036] Through the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented in hardware or by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0037] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A robot URDF model free combination and constraint management method, characterized in that, The method comprises: obtaining user demand data, determining a set of selectable components of a to-be-assembled robot according to the user demand data, the set of selectable components comprising a set of actuators and a set of sensors; calculating matching degrees of selectable components at each assembly position in the set of selectable components, and establishing a robot assembly scheme according to the matching degrees of the selectable components at each assembly position; predicting a predicted task execution effect of each robot assembly scheme, and selecting an optimal assembly scheme according to the predicted task execution effect; performing task simulation on the optimal assembly scheme based on a URDF model, establishing a trajectory constraint set of the robot according to a task simulation result, and controlling the robot to perform a task according to the trajectory constraint set.

2. The robot URDF model free combination and constraint management method according to claim 1, wherein, The set of selectable components of the to-be-assembled robot determined according to the user demand data comprises: establishing a natural language model, interpreting the user demand data based on the natural language model, and determining component function demand texts of each position corresponding to the user demand data according to an interpretation result; interpreting function description texts of components in a current robot component library based on the natural language model to obtain component function description texts of each component; matching the component function demand texts and the component function description texts to obtain selectable components at each position, and establishing the set of selectable components according to the selectable components at each position.

3. The robot URDF model free combination and constraint management method according to claim 2, wherein, The matching of the component function demand texts and the component function description texts to obtain the selectable components at each position comprises: performing vectorization processing on the component function demand texts and the component function description texts to obtain component function demand vectors and component function description vectors; calculating cosine similarity of the component function demand vectors and the component function description vectors, and selecting component function description vectors with cosine similarity greater than a first preset threshold; taking components corresponding to the selected component function description vectors as the selectable components at positions corresponding to the component function demand vectors, and obtaining the selectable components at all positions.

4. The robot URDF model free composition and constraint management method of claim 1, wherein, The establishment of the robot assembly scheme according to the matching degrees of the selectable components at each assembly position comprises: obtaining a component assembly point of any target selectable component, matching the component assembly point of the target selectable component with component assembly points of all selectable components at a corresponding assembly position; obtaining a matching assembly point quantity of matching components capable of matching the target selectable component at the assembly position, and counting assembly point types to which matching assembly points of each matching component belong; calculating matching degrees of the target selectable component and the matching components according to the assembly point types of the target selectable component and the matching components and the matching assembly point quantities of each assembly point type, selecting matching components with matching degrees greater than a second preset threshold than the target selectable component, and establishing the robot assembly scheme according to the selected matching components.

5. The robot URDF model free combination and constraint management method of claim 4, wherein, The calculation of the matching degrees of the target selectable component and the matching components according to the assembly point types of the target selectable component and the matching components and the matching assembly point quantities of each assembly point type comprises: determining the matching degrees of the target selectable component and the matching components according to a matching degree formula, the matching degree formula being, ， wherein, a matching degree of the target selectable component and the matching component, a first assembly point type importance weight, , , a preset standard required number of the target selectable component to the first assembly point type, a matching assembly point number of the matching component to the first assembly point type.

6. The robot URDF model free composition and constraint management method of claim 1, wherein, The prediction of the predicted task execution effect of each robot assembly scheme and the selection of the optimal assembly scheme according to the predicted task execution effect comprise: The robot performance prediction model is established, the input of the robot performance prediction model is the component parameters of each component after assembly, and the output is the predicted task execution effect; According to the historical robot task execution data, a training set is constructed, a loss function is established, and the robot performance prediction model is trained according to the training set and the loss function; The component parameters of the current assembly scheme are input into the trained robot performance prediction model, and the predicted task execution effect of the current assembly scheme is obtained.

7. The robot URDF model free combination and constraint management method of claim 6, wherein, The robot performance prediction model includes a component feature input module, a Transformer encoder module and an MLP module, the Transformer encoder module is used to capture the complex dependency relationship between the component parameters, and outputs the feature representation after the fusion of the component parameters, and the MLP module is used to map the fused feature representation to the final task execution effect.

8. The robot URDF model free combination and constraint management method of claim 6, wherein, The loss function is: ， wherein, is a loss function, is a predicted task success rate, is an actual task success rate, is an actual task completion time, is a predicted task completion time, , is a preset first weight and a preset second weight, .

9. The robot URDF model free composition and constraint management method of claim 1, wherein, The URDF model is used to simulate the task of the optimal assembly scheme, and the trajectory constraint set of the robot is established according to the task simulation result, including: According to the user demand data, the robot task data is determined, and the task constraint boundary is determined according to the robot task data; The URDF model of the optimal assembly scheme is established, and the URDF model of the optimal assembly scheme is imported into the physical simulation environment for task simulation; According to the robot task simulation result, the robot trajectory when executing the task is determined, when the robot trajectory when executing the task reaches the constraint boundary, the robot trajectory constraint is set, and the trajectory constraint set is established according to all the robot trajectory constraints.

10. A robot URDF model free combination and constraint management system, characterized in that, Including: The acquisition module is used to acquire user demand data, determine the selectable component set of the to-be-assembled robot according to the user demand data, and the selectable component set includes an actuator set and a sensor set; The matching module is used to calculate the matching degree of the selectable components in the assembly position of the selectable component set, and the robot assembly scheme is established according to the matching degree of the selectable components in the assembly position; The assembly module is used to predict the predicted task execution effect of each robot assembly scheme, and the optimal assembly scheme is selected according to the predicted task execution effect; The simulation module is used to simulate the task of the optimal assembly scheme based on the URDF model, and the trajectory constraint set of the robot is established according to the task simulation result, and the robot is controlled to execute the task according to the trajectory constraint set.

Images