A layout optimization method for a dual-arm robot based on teaching-based learning
By using teaching-based learning and DMP-based trajectory optimization to optimize the layout of the dual-arm robot, the time-consuming and labor-intensive problems in existing technologies are solved, achieving efficient layout and task execution.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG SCI-TECH UNIV
- Filing Date
- 2023-07-27
- Publication Date
- 2026-05-26
AI Technical Summary
Existing technologies consume a lot of time and computing power in optimizing the layout of dual-arm robots, and the global performance optimization has a limited impact on production tasks.
The teaching-learning method is adopted, and the expert trajectory is recorded through the vision system. The trajectory is learned and generalized using DMP. Combined with robot characteristics and task constraints, the layout index is optimized and the DMP parameters are adjusted to speed up task execution.
It improves the efficiency of layout optimization and task execution, reduces the workload of experts, and adapts to the specific needs of production tasks.
Smart Images
Figure CN116834013B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of dual-arm base optimization technology, specifically to a dual-arm robot layout optimization method based on teaching learning. Background Technology
[0002] The arrangement of the dual-arm base is fundamental to dual-arm manipulation. Current inventions and papers primarily focus on the positional calibration of the dual-arm layout, with less attention paid to the selection of the layout itself. Existing methods optimize metrics such as the product of the common workspace and the individual workspaces of each arm, rotation angle range, motion performance, and operability. This often requires global calculation and searching of the workspace, consuming significant time and computational resources. However, in practical applications of dual-arm robots in production, they often only operate along limited trajectories, such as dual-arm latte art, where global performance does not significantly impact their operational function. Summary of the Invention
[0003] In view of the shortcomings of the existing technology, the purpose of this invention is to provide a method for optimizing the layout of a dual-arm robot based on teaching learning.
[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for optimizing the layout of a dual-arm robot based on teaching and learning, characterized in that its steps are as follows:
[0005] Step 1: Teaching Recording: Use a vision system to read the hand trajectories of the expert performing the task in the workspace;
[0006] Step 2: DMP Learning: Use DMP to learn the trajectory of both hands;
[0007] Step 3: Selecting and generalizing common trajectory endpoints: Select the target points of the trajectories in the workspace, and generalize them to obtain commonly used trajectories in the space, reducing the workload of experts;
[0008] Step 4: Select the expected layout: Based on the robot's characteristics and task layout constraints, determine the set of undetermined robotic arm layouts;
[0009] Step 5: Design robot layout metrics: Based on existing metrics, establish a comprehensive metric that takes into account robot operational performance and the proportion of public workspace.
[0010] Step 6: Optimize the undetermined layout based on the indicators: Optimize the set with better performance from the undetermined layout set based on the indicators;
[0011] Step 7: Adjust learning algorithm parameters to generalize and execute faster task trajectories: Adjust DMP parameters to accelerate the evolution of generalized trajectories and obtain trajectories that can complete tasks faster. Due to the limitations of robot joints, new requirements are put forward for robot layout.
[0012] Step 8: Determine the layout method, and finally obtain the new layout method.
[0013] In some embodiments, according to step 2, the DMP model is established as follows:
[0014]
[0015] α z , is positive β z A constant, x is a variable, g is the target, and v is the velocity. Let τ be the acceleration, τ > 0 be the time constant, s be the number of time-independent stages, and the initial value be 1. α s Let f(s) be the convergence constant, and f(s) be the mandatory term.
[0016] In some embodiments, the mandatory term f(s) is:
[0017]
[0018] The mandatory terms include N Gaussian elements to ensure similarity of trajectories, where y0 is the starting point and w i It is the weight, ψ i It is a Gaussian kernel.
[0019] In some embodiments, the Gaussian kernel ψ i For: ψ i (s)=exp(-h i (sc i ) 2 ), w i Learn through LWR.
[0020] In some embodiments, according to steps 3-7, the calculation process is obtained by minimizing the objective function:
[0021] min{f t (s)-f(s)}
[0022] f(s) is the teaching trajectory, f t (s) is expressed as follows:
[0023]
[0024] By designing different initial and final configurations, the possible trajectories of the dual-arm robot task are obtained, providing a reference trajectory for the optimization of the dual-arm base.
[0025] In some embodiments, different base parameters are analyzed using a comprehensive evaluation of multiple indicators, as follows:
[0026] Maneuverability index, joint limits, and extended maneuverability;
[0027] Maneuverability index: Information about the robot's end effector's ability to unconditionally change its position and orientation; this information is inherently contained in the Jacobian matrix, specifically its singular values, denoted as,
[0028]
[0029] Joint Limit: When a joint variable approaches its limit, the degree of approach to the joint limit must tend towards infinity. The following function is used:
[0030]
[0031] Extended Maneuverability: To establish a maneuverability metric that incorporates joint limits and self-collision information, weight matrices must be computed. These matrices will help to construct an augmented Jacobian matrix that gives the extended maneuverability measure, using learned trajectories to avoid exploration and trial and error in all directions of motion.
[0032]
[0033]
[0034]
[0035] In some embodiments, the augmented Jacobian is calculated as follows:
[0036]
[0037] In some embodiments, the evaluation metric is set as follows:
[0038]
[0039] Compared with the prior art, the beneficial effects of the present invention are: based on teaching learning, it learns and generalizes the trajectory that the robot needs to run in the workspace, and by comprehensively considering the running efficiency and operational performance of the end-effector trajectory, it completes the layout optimization of the dual-arm robot base.
[0040] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. The embodiments of this application will provide a detailed description and understanding of the application. Attached Figure Description
[0041] Figure 1 This is a flowchart of the present invention. Detailed Implementation
[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0043] Please see Figure 1 This invention provides a technical solution: a teaching-based method for optimizing the layout of a dual-arm base, the specific implementation of which is as follows:
[0044] Step 1: Teaching record;
[0045] Use a vision system to read the hand movements of experts performing tasks in the workspace.
[0046] Step 2: DMP learning;
[0047] Use DMP to learn the trajectory of both hands.
[0048] Step 3: Select common trajectory endpoints and generalize the trajectory;
[0049] By selecting target points of trajectories in the workspace, commonly used trajectories within the space can be generalized, reducing the workload of experts.
[0050] Step 4: Select the desired layout method;
[0051] Based on the robot's characteristics and task layout constraints, a set of undetermined robotic arm layouts is determined.
[0052] Step 5: Design robot layout metrics;
[0053] Based on existing indicators, establish a comprehensive indicator that takes into account robot operation performance and the proportion of public workspace.
[0054] Step 6: Optimize the proposed layout based on the indicators;
[0055] Based on the indicators, select the best performing set from the set of undetermined layouts.
[0056] Step 7: Adjust the learning algorithm parameters to generalize and execute task trajectories faster;
[0057] Adjusting DMP parameters accelerates the evolution of generalized trajectories, resulting in trajectories that can complete tasks faster. Due to the limitations of robot joints, new requirements are placed on robot layout.
[0058] Step 8: Determine the layout method.
[0059] Finally, a new layout method was obtained.
[0060] The specific method for establishing the DMPs model is as follows:
[0061]
[0062] α z , is positive β z A constant, x is a variable, g is the target, and v is the velocity. Let τ be the acceleration, τ > 0 be the time constant, s be the number of time-independent stages, and the initial value be 1. α s The convergence constant is
[0063] The mandatory term f(s) is:
[0064]
[0065] The mandatory terms include N Gaussian elements to ensure similarity of trajectories, where y0 is the starting point and w i It is the weight, ψ i It is the Gaussian kernel, ψ i (s)=exp(-h i (sc i ) 2 ), w i You can learn through LWR.
[0066] The calculation process is obtained by minimizing the following objective function:
[0067] min{f t (s)-f(s)}
[0068] f(s) is the teaching trajectory, f t (s) are as follows:
[0069]
[0070] By designing different initial and final configurations, possible trajectories for the dual-arm robot task can be obtained, providing a reference trajectory for the optimization of the dual-arm base.
[0071] Similar to the configuration optimization across the entire space, we use a multi-index comprehensive evaluation and analysis for different base parameters. The specific indices are as follows:
[0072] Maneuverability index: Information about a robot's end effector's ability to unconditionally change its position and orientation. This information is inherently contained in the Jacobian matrix, particularly its singular values, denoted as...
[0073]
[0074] Joint Limit: For a joint variable to approach its limit, the degree of approach to the joint limit must be infinite. We use the following function...
[0075]
[0076] Extended Maneuverability: To establish a maneuverability metric that incorporates joint limits and self-collision information, weight matrices must be computed. These matrices will help to construct an augmented Jacobian matrix that gives the extended maneuverability measure. It is essentially dependent on the given motion; here, a learned trajectory is used to avoid exploration and trial and error in all directions of the motion.
[0077]
[0078]
[0079]
[0080] Finally, the augmented Jacobian can be calculated as
[0081]
[0082] Evaluation indicators can be set as follows:
[0083]
[0084] This technical solution proposes to use a teaching-learning method for layout optimization, thereby improving layout efficiency. Furthermore, by adjusting the parameters in the teaching-learning method, execution schemes with different running speeds can be obtained, thereby improving task execution efficiency.
[0085] Existing methods employ multiple metrics to analyze the entire workspace and optimize its layout, which consumes significant time and computing power. However, in reality, dual-arm robots used in production often perform similar, single tasks, and overall performance does not significantly impact their operation. This invention, based on teaching and learning, only requires optimizing the performance of the task and task-related areas. This not only enables faster layout optimization but also addresses the need for improved task efficiency.
[0086] The above embodiments merely illustrate several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.
[0087] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A layout optimization method for a dual-arm robot based on teaching-based learning, characterized in that: The steps are as follows: Step 1: Teaching Recording: Use a vision system to read the hand trajectories of the expert performing the task in the workspace; Step 2: DMP Learning: Use DMP to learn the trajectory of both hands; Step 3: Selecting and generalizing common trajectory endpoints: Select the target points of the trajectories in the workspace, and generalize them to obtain commonly used trajectories in the space, reducing the workload of experts; Step 4: Select the expected layout: Based on the robot's characteristics and task layout constraints, determine the set of undetermined robotic arm layouts; Step 5: Design robot layout metrics: Based on existing metrics such as maneuverability index, joint limits, and extended maneuverability, establish a comprehensive metric that takes into account robot operation performance and the proportion of public workspace. Step 6: Optimize the undetermined layout based on the indicators: Optimize the set with better performance from the undetermined layout set based on the indicators; Step 7: Adjust learning algorithm parameters to generalize and execute faster task trajectories: Adjust DMP parameters to accelerate the evolution of generalized trajectories and obtain trajectories that can complete tasks faster. Due to the limitations of robot joints, new requirements are put forward for robot layout. Step 8: Determine the layout method, and finally obtain the new layout method.
2. The method for optimizing the layout of a dual-arm robot based on teaching learning according to claim 1, characterized in that: Based on step 2, the DMP model is established as follows: , For positive integers, For variables, For the goal, For speed, For acceleration, The time constant, Time-independent number of stages, initial value 1. Let be the convergence constant. This is a mandatory item.
3. The method for optimizing the layout of a dual-arm robot based on teaching learning according to claim 2, characterized in that: Mandatory items for: The mandatory terms include N Gaussian elements to ensure trajectory similarity. It is the starting point. It's weight. It is a Gaussian kernel.
4. The method for optimizing the layout of a dual-arm robot based on teaching learning according to claim 3, characterized in that: Gaussian kernel for: , Learn through LWR.
5. The method for optimizing the layout of a dual-arm robot based on teaching learning according to claim 4, characterized in that: According to steps 3-7, the calculation process is obtained by minimizing the objective function: It is a teaching trajectory. The expression is as follows: By designing different initial and final configurations, the possible trajectories of the dual-arm robot task are obtained, providing a reference trajectory for the optimization of the dual-arm base.
6. The method for optimizing the layout of a dual-arm robot based on teaching learning according to claim 5, characterized in that: A comprehensive evaluation and analysis of different base parameters was conducted using multiple indicators, as follows: Maneuverability index, joint limits, and extended maneuverability; Maneuverability index: Information about the robot's end effector's ability to unconditionally change its position and orientation; this information is inherently contained in the Jacobian matrix, specifically its singular values, denoted as, Joint Limit: When a joint variable approaches its limit, the degree of approach to the joint limit must tend towards infinity. The following function is used: Extended Maneuverability: To establish a maneuverability metric that incorporates joint limits and self-collision information, weight matrices must be computed. These matrices will help to construct an augmented Jacobian matrix that gives the extended maneuverability measure, using learned trajectories to avoid exploration and trial and error in all directions of motion. 。 7. The method for optimizing the layout of a dual-arm robot based on teaching learning according to claim 6, characterized in that: The augmented Jacobian is calculated as follows: 。 8. The method for optimizing the layout of a dual-arm robot based on teaching learning according to claim 7, characterized in that: The evaluation indicators are set as follows: 。