Robot structure generative design and optimization method based on digital twinning

Abstract

The invention provides a robot structure generative design and optimization method based on digital twinning, and the method comprises the following steps: 1, building a mechanical arm structure generation model based on a deep neural network; the input of the mechanical arm structure generation model is task requirements, and the output is the rod length and the joint number of the mechanical arm; step 2, based on an improved particle swarm algorithm, optimizing the output of the mechanical arm structure generation model; wherein in the improved particle swarm algorithm, the inertia weight is an adaptive weight; and step 3, designing the robot and completing the construction of the robot in combination with the optimal solution output in the step 2. According to the invention, a robot can be generated based on a specific scene, and the performance evaluation indexes are diversified.

Description

Technical Field

[0001] The present invention belongs to the field of robotics technology, and in particular relates to a robot structure generative design and optimization method based on digital twins. Background Art

[0002] Robot structure design is one of the core areas of robotics development, involving multiple disciplines such as mechanical engineering, control engineering, and computer science. Its main goal is to design a robot structure that can meet specific task requirements and ensure its stability, flexibility, and efficiency in various environments.

[0003] The current state of the art in robot structural design and its shortcomings are mainly:

[0004] 1) Computer-aided design (CAD)

[0005] With the development of computer technology, CAD software has been widely used in robot structure design. CAD tools can help designers quickly create and modify design models, improve design efficiency and accuracy. However, CAD tools are mainly used in the design stage and lack automated optimization functions.

[0006] 2) Finite Element Analysis (FEA)

[0007] FEA technology is used to simulate and analyze the performance of robot structures under different loads and conditions. Through finite element analysis, designers can predict the stress, deformation, vibration and other characteristics of the structure to optimize the design. However, FEA usually requires a lot of computing resources and time, and relies on the designer's expertise.

[0008] 3) Generative Design

[0009] Generative design is an emerging design method that uses algorithms to generate multiple design solutions and selects the best solution through simulation and optimization. Generative design can explore a wider range of design spaces and discover innovative structures that are difficult to think of using traditional methods. However, the application of this technology in robot structural design is still in its early stages and faces challenges in algorithm complexity and computational cost.

[0010] 4) Digital Twin Technology

[0011] Digital twin technology enables real-time monitoring and optimization of robot structures by creating virtual models of physical objects. This technology can perform multiple simulations and tests in a virtual environment, reducing the need for physical prototypes. However, existing digital twin technology still needs to be improved in terms of intelligence and multi-physics field coupling simulation.

[0012] In addition, the current robot structure design still has the following problems:

[0013] 1) Reliance on manual design, which is labor-intensive. The current robot structure design process mostly relies on manual experience and trial and error, and designers need to manually adjust and optimize structural parameters. This method is not only time-consuming, but also requires more manpower as the design complexity increases, resulting in a waste of resources.

[0014] 2) Low design consistency. Since the design process relies on the designer’s personal experience and judgment, the consistency of the design results is low. Different designers may come up with different design solutions, making it difficult to ensure product consistency and quality in mass production.

[0015] 3) The optimization dimension is relatively single. Current design optimization mainly focuses on a single performance indicator, such as strength or weight, and rarely considers multi-dimensional comprehensive optimization. Existing technologies are difficult to optimize multiple performance indicators at the same time, resulting in the design solution may not meet multiple needs in actual applications.

[0016] 4) Application scenarios (task requirements) are relatively limited. The structure and parameters of the designed commercial robots are fixed, and it is difficult to fully guarantee the optimal comprehensive performance in terms of efficiency, cost, etc. under specific working conditions or application scenarios. Summary of the invention

[0017] The purpose of the present invention is to provide a robot structure generative design and optimization method based on digital twins to solve the above technical problems. The technical solution adopted is:

[0018] A robot structure generative design and optimization method based on digital twins includes the following steps:

[0019] Step 1: Establish a robotic arm structure generation model based on a deep neural network;

[0020] The input of the robot arm structure generation model is the task requirement, and its output is the rod length and the number of joints of the robot arm;

[0021] Step 2: Based on the improved particle swarm algorithm, the output of the robot arm structure generation model is optimized;

[0022] Among them, in the improved particle swarm algorithm, the inertia weight is an adaptive weight;

[0023] Step 3: Based on the optimal solution output in step 2, design the robot and complete its construction.

[0024] Preferably, in step 1, the specific process of establishing the mechanical arm structure generation model includes:

[0025] Step 1A, data collection and preprocessing: collect the rod lengths and joint numbers of different robotic arms, and calculate the working space of each robotic arm; mark the rod lengths and joint numbers according to different task requirements to form a structured data set;

[0026] Step 1B, deep learning model training: based on the data set, construct and train a deep learning neural network model until the model reaches a preset accuracy requirement;

[0027] Step 1C: Model saving and application: When the network model reaches the preset accuracy, the system automatically saves the model.

[0028] Preferably, in step 2, the expression of the inertia weight is:

[0029]

[0030] Among them, t-the current iteration number of the algorithm;

[0031] T-is the total number of iterations of the algorithm;

[0032] ω max - Inertia weight at the beginning of the algorithm;

[0033] ω min - Inertia weight at the end of the algorithm.

[0034] Preferably, the weight of the self-information is referenced when searching at iteration number t Its expression is:

[0035]

[0036] The weight of the reference group information when searching at iteration t Its expression is:

[0037]

[0038] The learning factor at the beginning of the algorithm;

[0039] is the learning factor at the beginning and end of the algorithm; i=1,2.

[0040] Compared with the prior art, the advantages of the present invention are:

[0041] 1. Combination of digital twins and generative design: The application of generative design for robots in actual scenarios is often unsatisfactory, and often faces problems such as collision, interference, or deviation between theory and actual effects. By introducing digital twin technology (integrated in the control system) for collision detection and obstacle avoidance path planning, the operation mode and effect of pose estimation and path planning can be simulated and optimized, and many problems that may be encountered in the hardware environment can be predicted and solved in advance, thereby improving accuracy, reliability, stability and safety, and realizing "using virtual to imitate the real, using virtual to reflect the real, using virtual to control the real, using virtual to predict the real, using virtual to improve the real, and the virtual and the real coexisting".

[0042] 2. Innovation in the method of generating structural parameters: Compared with the conventional reliance on assumptions or artificial experience to design robot structural parameters, a large number of robot structural parameter design samples are used, and a neural network is used as a prediction model to train the neural network to generate predictions of robot structural parameters. This avoids subjective assumptions and human intervention, saves manpower and time, and realizes the "self-organization" and "decentralization" of the design process.

[0043] 3. Universalization and promotion of application scenarios: Currently, commercial robots are generally only suitable for a single application scenario, and it is difficult to fully guarantee the optimal comprehensive performance in terms of efficiency, cost, etc. under specific working conditions or application scenarios. This robot structure generative design and optimization method based on digital twins can generate a robot based on a specific scenario, achieving the best comprehensive performance in the current scenario, while meeting the characteristics of lightweight, low energy consumption, and high precision.

[0044] 4. Diversification of performance evaluation indicators: Usually, the optimization indicators used for robot structures are relatively single, resulting in serious energy consumption of the robot and low motion trajectory tracking accuracy. This method quantitatively analyzes each key performance indicator (robot arm rod length, number of joints), determines a new optimization objective function (fitness function), further optimizes the robot arm layout, avoids waste of energy resources, and improves production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a flowchart of the generative design and optimization method of robot structure based on digital twin;

[0046] Figure 2 This is the structure diagram of the four-axis robot arm finally generated according to the optimal solution;

[0047] Figure 3 This is the optimization result graph.

[0048] Among them, 1-the first joint of the robot arm's moving pair; 2-the second joint of the robot arm's rotating pair; 3-the third joint of the robot arm's rotating pair; 4-the fourth joint of the robot arm's rotating pair; 5-the end gripper of the robot arm. DETAILED DESCRIPTION

[0049] The following is a more detailed description of the robot structure generative design and optimization method based on digital twins of the present invention in conjunction with a schematic diagram, which shows a preferred embodiment of the present invention. It should be understood that those skilled in the art can modify the present invention described herein and still achieve the beneficial effects of the present invention. Therefore, the following description should be understood as being widely known to those skilled in the art and not as a limitation of the present invention.

[0050] like Figure 1 As shown, a robot structure generative design and optimization method based on digital twins includes the following steps:

[0051] 1) Generative design of robot structure based on digital twin

[0052] This technical route realizes the automatic generation and optimization of the structural parameters of the robotic arm through deep learning neural networks.

[0053] That is, step 1, based on the deep neural network, a robotic arm structure generation model is established.

[0054] The input of the robotic arm structure generation model is the task requirement, and its output is the rod length and number of joints of the robotic arm; the robot is the robotic arm.

[0055] Step 1 specifically includes the following steps:

[0056] Step 1A, data collection and preprocessing: collect the rod lengths and joint numbers of different robotic arms, and calculate the workspace of each robotic arm; the same workspace may correspond to different task requirements. According to the different task requirements of the workspace, mark the rod lengths and joint numbers to form a structured data set.

[0057] Among them, the workspace is calculated to evaluate the reachability of the robot arm and verify whether the robot arm workspace meets the task requirements of the work scenario, that is, whether it can reach the longest distance required for the work.

[0058] Step 1B: Deep learning model training.

[0059] Based on the data set, a deep learning neural network model is built and trained. The input of the model is the task requirements, and the output is the rod length and number of joints of the robot arm. Through continuous iterative training, the model parameters are optimized until the model reaches the preset accuracy requirements.

[0060] Step 1C: Save and apply the model.

[0061] When the network model reaches the preset accuracy, the system automatically saves the model.

[0062] This model can be used to predict and generate the structural parameters of the robotic arm that meet specific task requirements.

[0063] The pseudo code is as follows:

[0064]

[0065] Among them, one task requirement corresponds to one work label.

[0066] 2) Robot structure optimization based on digital twin

[0067] That is, in step 2, based on the improved particle swarm algorithm, the output of the robot arm structure generation model is optimized.

[0068] Step 2 specifically includes the following steps:

[0069] (1) Initialization: Randomly generate the same number of particles as the predetermined dimension in the search space and determine their initial positions and velocities.

[0070] Assume that in the D-dimensional search space, a particle swarm is composed of N particles, and the current iteration number of the algorithm is t, then the position vector of the i-th particle is:

[0071]

[0072] The velocity vector of the i-th particle is:

[0073]

[0074] (2) Evaluation: Calculate the fitness value of each particle and determine the quality of the solution.

[0075] For this study, the time taken by the robot to reach the target position from the initial position, the overall mass of the robot, and the energy consumed by the robot movement were used as evaluation parameters.

[0076] Among them, the fitness function is:

[0077]

[0078] T represents time, M represents mass, E represents energy, and the denominator represents the maximum time, mass, and energy before optimization. The optimized division is normalized. γ1~γ3 are weights. The weights are set based on the importance of the three indicators. The specific values ​​are set to 6, 2, and 2.

[0079] That is, the fitness function includes and These three parts.

[0080] (3) Update individual extreme value: If the current fitness of a particle is higher than its individual extreme value, set the individual extreme value of the particle to its current position;

[0081] The velocity of the i-th particle is updated as follows:

[0082]

[0083] is the inertia part, which represents the tendency of the particle to maintain its current state of motion;

[0084] The individual cognitive part represents the tendency of the particle to move to its best position according to its previous search results;

[0085] It is the group cognition part, which represents the tendency of particles to move toward the global optimal position according to the group's previous search experience.

[0086] calculate After that, the position of the i-th particle is updated as follows:

[0087]

[0088] Among them, r 1i and r 2i Uniformly distributed in the range [0,1], it is a random number about particle i, used to increase the randomness of the search.

[0089] ω t It is the inertia weight of the particle corresponding to the current number of iterations, which determines the degree of inheritance of the particle velocity. It gradually decreases with the increase of the number of iterations t to speed up the iteration speed of the algorithm.

[0090]

[0091] and is the learning factor, which represents the weight of the particle referring to its own information and group information when searching for iteration t. The individual learning factor As the number of iterations t increases, the social learning factor decreases. As the number of iterations t increases, it gradually increases:

[0092]

[0093] Among them, ω max and ω min are the inertia weights at the beginning and end of the algorithm, and

[0094] (i=1,2) is the learning factor at the beginning and end of the algorithm, and T is the total number of iterations of the algorithm. Obviously, as the number of iterations increases, the above parameters increase or decrease in a linear relationship.

[0095] (4) Update the global extreme value: If the individual extreme value of the current particle is better than the global extreme value of the entire particle group, the global extreme value is updated to the individual extreme value of the particle.

[0096] (5) Update inertia weight, velocity and position: First, update the inertia weight of the particle, and then adjust the velocity and position of each particle according to the individual extreme value and the global extreme value.

[0097] By repeating steps (2) to (5) until the number of iterations is reached, the global optimal solution can be finally obtained.

[0098] The pseudo code is as follows:

[0099]

[0100] In summary, in step 2, the length of the robotic arm and the number of joints are used as the optimization objectives of the algorithm. Combined with the robot evaluation indicators, the running time, energy consumption and the overall quality of the robotic arm are designed as the algorithm evaluation function (fitness function), so as to achieve the purpose of optimizing the performance of the generative robot.

[0101] like Figure 3 As shown in the figure, after algorithm optimization, the three evaluation indicators of the robot's operating time, total mass and energy consumption have all achieved good optimization effects and the optimization results are in line with the actual scenario requirements, ensuring that the robot can shorten the working time, reduce the mass and reduce the energy consumption while completing the task.

[0102] Figure 3 In , the horizontal axis is the number of iterations, and the vertical axis is the value of each part of the fitness function.

[0103] Step 3: Based on the optimal solution output in step 2, design the robot and complete its construction.

[0104] According to the scene requirements of the sorting task, this robotic arm is generated by the above-mentioned generative system and finally completes the construction and debugging of the robotic arm, and is finally applied to relevant scenes. The specific implementation method is as follows:

[0105] 1. Object detection and pose estimation of sorting targets:

[0106] (1) Use the YOLOv8 algorithm for target detection, including target recognition and location acquisition.

[0107] (2) Target recognition and pose estimation based on depth camera and geometric features.

[0108] (3) Use image processing techniques (such as Canny edge detection and Hough transform) to assist in pose estimation.

[0109] 2. Robot structure generation and optimization:

[0110] (1) Using the deep learning model and inputting the scene requirements, we obtain the basic structure of a sorting robot that can complete the sorting within a radius of 700 mm to 800 mm and a vertical height difference of 800 mm. The robot can grasp targets of different shapes, including cylinders, cubes, etc., with a load not exceeding 1 kg.

[0111] (2) When the model reaches the preset accuracy, the model is automatically saved. The model is the robot structure parameters that meet the specific task requirements.

[0112] (3) The physical model of the robot's basic structure is optimized by improving the algorithm. The algorithm adjusts the robot structure in a targeted manner through iteration, reorganizes the joints, adjusts the rod length, and simulates the robot's operating scenario to obtain the final robot structure model and parameters.

[0113] 3. Construction of Generative Robotic Arm Digital Twin System:

[0114] (1) Based on the generated robotic arm structure and its parameters, design the robot and complete its construction.

[0115] (2) Integrate target detection and pose estimation to build a complete fully automatic robot system.

[0116] (3) Realize the control of the robot and combine it with the digital twin system (integrated into the control system).

[0117] (4) Run the robot and digital twin system to complete the scheduled sorting tasks.

[0118] like Figure 2 As shown, the robot includes:

[0119] Robotic arm moving sub-level 1;

[0120] The second joint 2 of the mechanical arm rotation pair is connected to the first joint 1 of the mechanical arm movement pair and can move up and down with it;

[0121] The third joint 3 of the revolving pair of the robot arm is connected with the fourth joint 4 of the revolving pair of the robot arm and the second joint 2 of the revolving pair of the robot arm through rods.

[0122] The gripper 5 at the end of the robot arm is connected to the fourth joint 4 of the robot arm revolute pair.

[0123] The controller is connected with the signal of the first joint 1 of the moving pair of the robot arm.

[0124] The digital twin system integrates target detection and pose estimation and is connected to the controller signal.

[0125] The above is only a preferred embodiment of the present invention and does not limit the present invention in any way. Any technician in the relevant technical field, without departing from the scope of the technical solution of the present invention, makes any form of equivalent replacement or modification to the technical solution and technical content disclosed in the present invention, which does not depart from the content of the technical solution of the present invention and still falls within the protection scope of the present invention.

Claims

1. A robot structure generative design and optimization method based on digital twin, characterized in that: The following steps are involved: Step 1: Establish a robotic arm structure generation model based on a deep neural network; The input of the robot arm structure generation model is the task requirement, and its output is the rod length and the number of joints of the robot arm; Step 2: Based on the improved particle swarm algorithm, the output of the robot arm structure generation model is optimized; Among them, in the improved particle swarm algorithm, the inertia weight is an adaptive weight; Step 3: Based on the optimal solution output in step 2, design the robot and complete its construction.

2. The robot structure generative design and optimization method based on digital twin according to claim 1, characterized in that: In step 1, the specific process of establishing the robotic arm structure generation model includes: Step 1A, data collection and preprocessing: collect the rod lengths and joint numbers of different robotic arms, and calculate the working space of each robotic arm; mark the rod lengths and joint numbers according to different task requirements to form a structured data set; Step 1B, deep learning model training: based on the data set, construct and train a deep learning neural network model until the model reaches a preset accuracy requirement; Step 1C: Model saving and application: When the network model reaches the preset accuracy, the system automatically saves the model.

3. The robot structure generative design and optimization method based on digital twin according to claim 1, characterized in that: In step 2, the expression of the inertia weight is: Among them, t-the current iteration number of the algorithm; T-is the total number of iterations of the algorithm; ω max - Inertia weight at the beginning of the algorithm; ω min - Inertia weight at the end of the algorithm.

4. The robot structure generative design and optimization method based on digital twin according to claim 3 is characterized in that: The weight of the self-information is referenced when searching at iteration t Its expression is: The weight of the reference group information when searching at iteration t Its expression is: The learning factor at the beginning of the algorithm; is the learning factor at the beginning and end of the algorithm; i=1,2.

Images