A PID-based servo drive control method and system

Abstract

The application provides a PID-based servo drive control method and system, and relates to the technical field of servo drive control.The method comprises the following steps: acquiring robot capture operation videos of small arthropods in a salt marsh under different mechanical claw clamping forces; determining a lower limit mechanical claw clamping force for preventing falling and an upper limit mechanical claw clamping force for protecting life based on operation information under each mechanical claw clamping force; determining a plurality of first mechanical claw clamping forces based on the lower limit mechanical claw clamping force for preventing falling, the upper limit mechanical claw clamping force for protecting life and the operation information under each mechanical claw clamping force; determining a target mechanical claw clamping force based on robot capture operation videos of small arthropods in a salt marsh under the plurality of first mechanical claw clamping forces; and performing PID servo drive control on the robot based on the target mechanical claw clamping force.The method can accurately determine the best adaptive mechanical claw clamping force of the robot for capturing small arthropods in a salt marsh.

Description

Technical Field

[0001] This invention relates to the field of servo drive control technology, and specifically to a servo drive control method and system based on PID. Background Technology

[0002] Small arthropods in salt marshes are a core component of wetland ecosystems, playing an irreplaceable role in maintaining biodiversity, promoting nutrient cycling, and ensuring ecological balance. Their sampling and ecological monitoring are crucial foundations for wetland ecological research. However, the unique environment of salt marshes—with its soft substrate and complex topography—combined with the small size, fragile shells, and agile movements of these small arthropods, presents numerous challenges to accurate capture. Traditional manual capture methods are inefficient and prone to causing mechanical damage to the organisms, affecting the reliability of subsequent research data. Conventional robotic capture equipment often uses fixed parameters or empirically adjusted gripping forces, lacking specific adaptive designs, making it difficult to balance the dual requirements of stable capture to prevent detachment and protecting the organism's survival and morphological integrity. Insufficient gripping force leads to inadequate capture stability, significantly reducing the success rate and potentially causing escape and sample collection failure. Excessive gripping force, on the other hand, can cause irreversible damage to the structure of small arthropods, inducing injury, death, or morphological deformities, thus interfering with the objectivity and accuracy of subsequent research data.

[0003] Therefore, accurately determining the optimal gripping force of the mechanical claw for robots to capture small arthropods in salt marshes is a problem that urgently needs to be solved. Summary of the Invention

[0004] The main technical problem this invention addresses is how to accurately determine the optimal gripping force of the mechanical claw for a robot to capture small arthropods in salt marshes.

[0005] According to a first aspect, the present invention provides a PID-based servo drive control method, comprising: acquiring robot capture operation videos under different mechanical claw gripping forces for small arthropods in salt marshes; determining operational information for each mechanical claw gripping force based on the robot capture operation videos under different mechanical claw gripping forces for small arthropods in salt marshes using a capture performance determination model; determining a lower limit mechanical claw gripping force for preventing detachment and an upper limit mechanical claw gripping force for protecting life based on the operational information for each mechanical claw gripping force; determining multiple first mechanical claw gripping forces based on the lower limit mechanical claw gripping force for preventing detachment, the upper limit mechanical claw gripping force for protecting life, and the operational information for each mechanical claw gripping force; acquiring robot capture operation videos under the multiple first mechanical claw gripping forces for capturing small arthropods in salt marshes; determining a target mechanical claw gripping force based on the robot capture operation videos under the multiple first mechanical claw gripping forces for capturing small arthropods in salt marshes; and performing PID servo drive control on the robot based on the target mechanical claw gripping force to complete the capture operation of small arthropods in salt marshes.

[0006] In one possible implementation, determining the target mechanical gripper gripping force based on the robot's operation video capturing small arthropods in salt marsh wetlands under the gripping forces of the plurality of first mechanical grippers includes: constructing a mechanical gripper gripping map, wherein the mechanical gripper gripping map includes a plurality of first mechanical gripper gripping force nodes and multiple edges between the plurality of first mechanical gripper gripping force nodes, the plurality of first mechanical gripper gripping force nodes are connected in ascending order of mechanical gripping force, and the node features of each first mechanical gripper gripping force node include the robot's operation video capturing small arthropods in salt marsh wetlands under each first mechanical gripping force, and the edge is the mechanical gripping force between two first mechanical gripper gripping force nodes. The force difference of the gripping force of the mechanical claw; the mechanical claw gripping map is processed based on a graph neural network to determine the gripping force of multiple second mechanical claws; based on the robot capture operation video of small arthropods in salt marsh wetlands under different mechanical claw gripping forces, the robot capture operation video of small arthropods in salt marsh wetlands under multiple first mechanical claw gripping forces, and the multiple second mechanical claw gripping forces, a robot capture operation simulation video of small arthropods in salt marsh wetlands is generated for each second mechanical claw gripping force; the target mechanical claw gripping force is determined based on the robot capture operation simulation video of small arthropods in salt marsh wetlands under each second mechanical claw gripping force.

[0007] In one possible implementation, the operational information under the gripping force of each mechanical claw includes the survival status information and shape change information of the target organism.

[0008] In one possible implementation, the input to the graph neural network is the mechanical gripper pattern, and the output of the graph neural network is the gripping force of multiple second mechanical grippers.

[0009] According to a second aspect, the present invention provides a servo drive control system based on PID, comprising: a first acquisition module, configured to acquire robot capture operation videos under different mechanical claw gripping forces of small arthropods in salt marshes; an operation information determination module, configured to determine operation information under each mechanical claw gripping force based on the robot capture operation videos under different mechanical claw gripping forces of small arthropods in salt marshes using a capture performance determination model; a gripping force range determination module, configured to determine the lower limit of anti-detachment mechanical claw gripping force and the upper limit of life-protection mechanical claw gripping force based on the operation information under each mechanical claw gripping force; and a first gripping force determination module, configured to... The system comprises: a first mechanical claw gripping force determined by the lower limit gripping force of the anti-detachment mechanical claw, the upper limit gripping force of the life-protection mechanical claw, and the operational information under each mechanical claw gripping force; a second acquisition module for acquiring video footage of the robot capturing small arthropods in salt marshes under the multiple first mechanical claw gripping forces; a target gripping force determination module for determining a target mechanical claw gripping force based on the video footage of the robot capturing small arthropods in salt marshes under the multiple first mechanical claw gripping forces; and a control execution module for performing PID servo drive control on the robot based on the target mechanical claw gripping force to complete the capture operation of small arthropods in salt marshes.

[0010] In one possible implementation, the target gripping force determination module is further configured to: construct a mechanical gripper gripping map, the mechanical gripper gripping map including multiple first mechanical gripper gripping force nodes and multiple edges between the multiple first mechanical gripper gripping force nodes, the multiple first mechanical gripper gripping force nodes being connected in ascending order of gripping force, the node features of each first mechanical gripper gripping force node including video footage of the robot capturing small arthropods in salt marsh wetlands under each first mechanical gripping force, and the edge representing the force difference of the gripping force between two first mechanical gripper gripping force nodes; based on a graph neural network. The mechanical claw gripping map is processed to determine multiple second mechanical claw gripping forces; based on the robot capture operation videos of small arthropods in salt marshes under different mechanical claw gripping forces, the robot capture operation videos of small arthropods in salt marshes under multiple first mechanical claw gripping forces, and the multiple second mechanical claw gripping forces, a robot capture operation simulation video of small arthropods in salt marshes is generated for each second mechanical claw gripping force; the target mechanical claw gripping force is determined based on the robot capture operation simulation video of small arthropods in salt marshes under each second mechanical claw gripping force.

[0011] In one possible implementation, the operational information under the gripping force of each mechanical claw includes the survival status information and shape change information of the target organism.

[0012] In one possible implementation, the input to the graph neural network is the mechanical gripper pattern, and the output of the graph neural network is the gripping force of multiple second mechanical grippers.

[0013] According to a third aspect, embodiments of the present invention provide an electronic device, including: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method comprising: acquiring robot capture operation videos of small arthropods in salt marshes under different mechanical claw gripping forces; determining operational information for each mechanical claw gripping force based on the robot capture operation videos of small arthropods in salt marshes under different mechanical claw gripping forces using a capture performance determination model; and determining anti-detachment measures based on the operational information for each mechanical claw gripping force. The robot employs a limited gripping force and an upper limit gripping force for protection. Based on the lower limit gripping force for preventing detachment, the upper limit gripping force for protection, and operational information under each gripping force, multiple first gripping forces are determined. Videos of the robot capturing small arthropods in salt marshes under these multiple first gripping forces are acquired. A target gripping force is determined based on these videos. PID servo drive control is applied to the robot based on the target gripping force to complete the capture operation of small arthropods in salt marshes.

[0014] According to the fourth aspect, this embodiment provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the program implements the aforementioned PID-based servo drive control method. The method includes: acquiring robot capture operation videos under different mechanical claw gripping forces of small arthropods in salt marshes; determining operational information for each mechanical claw gripping force using a capture performance determination model based on the robot capture operation videos under different mechanical claw gripping forces of small arthropods in salt marshes; and determining the lower limit of anti-detachment mechanical claw gripping force based on the operational information for each mechanical claw gripping force. The robot employs the following methods: 1) Upper limit gripping force of the robotic claw; 2) Determine multiple first gripping forces of the robotic claw based on the lower limit gripping force of the robotic claw, the upper limit gripping force of the robotic claw, and operational information under each gripping force; 3) Acquire video footage of the robot capturing small arthropods in salt marshes under the multiple first gripping forces; 4) Determine the target gripping force of the robotic claw based on the video footage of the robot capturing small arthropods in salt marshes under the multiple first gripping forces; 5) Perform PID servo drive control on the robot based on the target gripping force to complete the capture operation of small arthropods in salt marshes.

[0015] This invention provides a servo drive control method and system based on PID control. The method includes acquiring robot capture operation videos under different gripping forces of mechanical claws on small arthropods in salt marshes; determining operational information for each gripping force using a capture performance determination model based on the robot capture operation videos under different gripping forces of mechanical claws on small arthropods in salt marshes; determining a lower limit gripping force for preventing detachment and an upper limit gripping force for protecting life based on the operational information for each gripping force; and determining the lower limit gripping force for preventing detachment and the upper limit gripping force for protecting life based on the lower limit gripping force for preventing detachment and the upper limit gripping force for protecting life. The method involves determining multiple first mechanical claw gripping forces based on the operational information under each mechanical claw gripping force; acquiring video footage of the robot capturing small arthropods in salt marshes under these first mechanical claw gripping forces; determining a target mechanical claw gripping force based on the video footage; and executing PID servo drive control on the robot based on the target mechanical claw gripping force to complete the capture operation of small arthropods in salt marshes. This method can accurately determine the optimal mechanical claw gripping force for the robot to capture small arthropods in salt marshes. Attached Figure Description

[0016] Figure 1 A flowchart illustrating a PID-based servo drive control method provided in an embodiment of the present invention;

[0017] Figure 2 A schematic diagram of a robot provided for an embodiment of the present invention;

[0018] Figure 3 This is a schematic diagram of a process for determining the gripping force of a target mechanical claw, provided in an embodiment of the present invention.

[0019] Figure 4 A schematic diagram of a PID-based servo drive control system provided in an embodiment of the present invention; Detailed Implementation

[0020] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to the present invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0021] In this embodiment of the invention, the following are provided: Figure 1 The illustrated servo drive control method based on PID includes steps S1 to S7:

[0022] Step S1: Obtain video footage of the robot capturing and operating under different mechanical claw gripping forces on small arthropods in the salt marsh wetland.

[0023] The video of the robot capturing small arthropods in salt marsh wetlands under different gripping forces of the mechanical claws is recorded by a camera video acquisition device installed on the robot as it captures small arthropods in the salt marsh wetland environment with various gripping forces. Figure 2 This is a schematic diagram of a robot provided in an embodiment of the present invention.

[0024] Robot capture operation videos can record the entire process of robot capture actions under different gripping forces, such as the core steps of the mechanical claw approaching, contacting, and gripping the organism, as well as the real-time state changes of small arthropods in salt marshes during the capture process.

[0025] Step S2: Based on the robot capture operation video of small arthropods in the salt marsh wetland under different mechanical claw gripping forces, the capture performance determination model is used to determine the operation information under each mechanical claw gripping force.

[0026] The capture performance determination model is a Transformer model. The input to the capture performance determination model is a video of the robot capturing small arthropods in the salt marsh wetland under different mechanical claw gripping forces, and the output of the capture performance determination model is the operation information under each mechanical claw gripping force.

[0027] The Transformer model is a deep learning model that employs a self-attention mechanism. It can process sequential data in parallel and effectively capture long-distance dependencies between elements in the sequence. The core structure of the Transformer model consists of an encoder and a decoder. The encoder transforms the input sequence into a deep feature representation, while the decoder generates the target output based on these features. The Transformer's self-attention mechanism allows the model to simultaneously weigh and reference information from all other points in the sequence when processing information at a particular time point, thus establishing a precise contextual understanding of time-series data.

[0028] The operational information for each gripping force of the robotic claw was obtained by analyzing robot capture operation videos of small arthropods in salt marshes under different gripping forces using a capture performance determination model. This information corresponds to the capture effectiveness at each gripping force. The operational information for each gripping force includes the survival status and shape changes of the target organism.

[0029] The survival status information of the target organism is the life status information of the small arthropod organism in the salt marsh wetland after being captured by the mechanical claw under the corresponding gripping force.

[0030] The survival status information of the target organisms includes whether the captured small arthropods in the salt marsh wetlands are alive, their activity index under survival status, and whether they have been injured or have experienced stress.

[0031] The live activity index is a numerical indicator that quantifies the limb movement, locomotion, and responsiveness to external stimuli of a captured small arthropod from a salt marsh wetland after being gripped by a mechanical claw. The live activity index ranges from 0 to 10, with higher values ​​indicating greater activity.

[0032] Shape change information quantifies the degree of abnormal changes in the body shape of captured small arthropods from salt marshes when subjected to the gripping force of mechanical claws. Shape change information includes the compression ratio of the captured small arthropod's body, the depth of indentations at specific locations, the area of ​​shell damage, the bending angle of limbs, and the degree of fracture coefficient.

[0033] The body compression ratio is calculated based on the natural body volume of the captured small arthropod from the salt marsh wetland, representing the percentage reduction in volume after clamping.

[0034] The depth of indentation in a specific location includes the deformation depth value in millimeters for key areas such as the abdomen and limb joints.

[0035] The limb bending angle is calculated based on the normal limb angle of the captured small arthropod in the salt marsh wetland, and the deviation angle of the limb after clamping is measured in degrees.

[0036] The fracture severity coefficient is a grading index used to quantify the degree of limb or shell fracture of captured small arthropods in salt marshes under the grip of mechanical claws. The fracture severity coefficient is quantified on a scale of 0 to 5, with 0 representing no fracture and 5 representing complete fracture.

[0037] The video footage of a robot capturing small arthropods in salt marshes under varying gripping forces of its mechanical claws forms a time series. This series includes the dynamic behavior of the captured creatures, the trajectory of the mechanical claws, and all the visual details of their interaction. These visual details provide a direct basis for evaluating the capture effectiveness. For example, key events such as minute body deformations, the frequency and amplitude of struggling movements, and whether the creatures were successfully gripped are all recorded in the video as pixel data.

[0038] The Transformer model, through its self-attention mechanism, can process video footage of a robot capturing small arthropods in salt marshes under varying gripping forces. The Transformer model decomposes the video into a continuous sequence of image frames and extracts spatial features from each frame. These feature sequences are then fed into the Transformer's encoder. Inside the encoder, the self-attention mechanism calculates the association weights between each frame and all other frames in the sequence. This allows the model to simultaneously consider key segments throughout the process, such as the natural state of the small arthropod before capture, the impact and deformation at the moment of capture, and the organism's reaction after capture. For example, to determine the survival status of the captured target small arthropod, the model can focus on whether there are continuous, regular patterns of limb movement in the post-capture video frame sequence and compare them with the activity patterns before capture. To determine shape change information, the model can precisely compare the organism's contour at the moment of capture with its contour before capture at the pixel level and quantify the body's compression rate and the degree of local deformation.

[0039] Step S3: Determine the lower limit of anti-detachment mechanical claw clamping force and the upper limit of life-saving mechanical claw clamping force based on the operation information under the clamping force of each mechanical claw.

[0040] In some embodiments, a clamping force model can be used to determine the lower limit clamping force of the anti-detachment mechanical claw and the upper limit clamping force of the life-protection mechanical claw. The clamping force model is a deep neural network model. The input to the clamping force model is the operational information under each mechanical claw clamping force, and the output of the clamping force model is the lower limit clamping force of the anti-detachment mechanical claw and the upper limit clamping force of the life-protection mechanical claw.

[0041] Deep neural network models include Deep Neural Networks (DNNs), which are computational models composed of multiple processing layers (neuron layers). Deep neural networks can learn complex nonlinear relationships in data by simulating the connections between neurons in the human brain. Deep neural network models can automatically extract and learn features from large amounts of input data and establish a precise mapping from input to output through layer-by-layer abstraction. The structure of a deep neural network includes an input layer, multiple hidden layers, and an output layer. By continuously adjusting the weights of the connections between layers during training, deep neural networks can approximate arbitrarily complex functions.

[0042] The minimum gripping force of the anti-escape mechanical claw is determined by the gripping force model. It is the minimum gripping force required by the mechanical claw to ensure stable capture of small arthropods in salt marshes and to prevent small arthropods from escaping during capture or transfer.

[0043] Using mechanical grippers with a gripping force below the lower limit of anti-detachment will significantly increase the risk of capture failure or instability of the target small arthropods in the salt marsh wetlands.

[0044] The maximum gripping force of the biosafety mechanical claw is determined by the gripping force model. It is the maximum gripping force that the mechanical claw can apply under the premise of ensuring the survival of small arthropods in salt marshes and preventing irreversible damage to their body shape.

[0045] Exceeding the protection limit of mechanical claws will cause irreversible physical damage to the captured small arthropods in the salt marsh wetlands, significantly reduce their survival rate, or even directly cause their death, thus failing to meet the protection needs of the captured organisms.

[0046] The operational information for each gripping force of the robotic claw includes the survival status and shape change information of small arthropods in the salt marsh wetlands corresponding to different gripping forces. The survival status and shape change information can intuitively reflect the degree of impact of gripping force on the organisms, thus enabling the model to find a critical force value that balances the need to prevent escape and the need to protect the captured small arthropods in the salt marsh wetlands.

[0047] Deep neural networks can receive operational information under each gripping force of a robotic claw as input, using the gripping force as an input feature, and then learn the shape change information and survival status information of the captured target small arthropods in salt marshes as output targets. The deep neural network can learn a function and predict the degree of shape change under a given force. After training, the model can grasp the comprehensive influence of gripping force on the capture effect. The model can set a minimum stable gripping standard to ensure that the captured target small arthropods in salt marshes do not detach, such as no displacement or loosening during gripping. Then, using the learned relationship between force and gripping stability, it finds the minimum force value that meets this stable gripping standard; this value is the lower limit gripping force of the robotic claw to prevent detachment. In addition, the model can set an acceptable maximum shape change threshold and a minimum survival status standard. The model can use the learned relationship between force and damage to find the critical force value that is about to exceed these safety thresholds; this critical force value is determined as the upper limit gripping force of the robotic claw to protect the survivors.

[0048] Step S4: Determine multiple first mechanical claw clamping forces based on the lower limit clamping force of the anti-detachment mechanical claw, the upper limit clamping force of the life-protecting mechanical claw, and the operation information under each mechanical claw clamping force.

[0049] In some embodiments, a first force determination model can be used to determine the gripping forces of multiple first robotic grippers. The first force determination model is a generative adversarial network (GAN). The inputs to the first force determination model are the lower limit gripping force of the anti-detachment robotic gripper, the upper limit gripping force of the life-saving robotic gripper, and operational information under each gripping force of the robotic gripper. The output of the first force determination model is the multiple gripping forces of the first robotic grippers.

[0050] Generative Adversarial Networks (GANs) are deep learning models consisting of a generator and a discriminator. GANs achieve data generation or decision optimization through adversarial training between these two networks. The generator produces candidate data that meets the requirements based on input constraints and data distribution characteristics. The discriminator distinguishes and evaluates the candidate data output by the generator from real, valid data, and provides an evaluation score. During training, the generator continuously optimizes its parameters to generate results that better meet the discriminator's evaluation criteria, while the discriminator continuously improves its discriminative ability. The generator and discriminator compete and reinforce each other, ultimately enabling the generator to output high-quality results that meet the constraints.

[0051] Multiple first mechanical gripper gripping forces are a new and exploratory set of mechanical gripper gripping force candidate values ​​generated by the first force determination model within a safe and effective force range consisting of the lower limit mechanical gripper gripping force for preventing detachment and the upper limit mechanical gripper gripping force for protecting life.

[0052] The lower limit of anti-detachment mechanical claw gripping force and the upper limit of life-protection mechanical claw gripping force can define the effective range of gripping force. The operation information under each mechanical claw gripping force provides actual performance data corresponding to different forces, including the survival status and shape changes of the captured target small arthropods in the salt marsh wetland. This data enables the candidate gripping force values ​​generated by the generator to have better overall performance within the effective range, and also provides the discriminator with a standard for evaluating the quality of candidate gripping force values.

[0053] The generator of the generative adversarial network (GAN) can use the lower limit of anti-detachment gripping force and the upper limit of life-protection gripping force as rigid constraints, strictly ensuring that the generated candidate forces are always within a safe and effective range. Simultaneously, the generator can deeply analyze the actual performance data, such as biological survival status and shape change information, contained in the operational information under each gripping force. This allows for precise capture of the performance distribution patterns of different gripping forces in the dimensions of stable gripping and biological protection, thereby generating candidate gripping forces within the effective range that are exploratory and better meet the requirements of excellent biological survival status and minimal shape change. The discriminator then uses the operational information under each gripping force as a real reference standard to evaluate the candidate forces output by the generator, determining whether they meet the basic requirements and comprehensive performance potential of the stable gripping standard corresponding to the lower limit of anti-detachment and the biological protection standard corresponding to the upper limit of life-protection. During adversarial training, the generator can continuously optimize its parameters to generate candidate strengths that better meet performance requirements, while the discriminator continuously improves its ability to distinguish between real high-quality data and candidate data. The two promote each other through dynamic iteration, ultimately enabling the generator to output multiple first-hand gripping forces within a safe range that combine exploratory value and comprehensive performance advantages.

[0054] Step S5: Obtain multiple videos of the robot capturing small arthropods in salt marsh wetlands under the gripping force of the first mechanical claw.

[0055] The video of the robot capturing small arthropods in salt marsh wetlands under the gripping force of multiple first mechanical claws is recorded by the camera video acquisition device on the robot when the robot captures small arthropods in the salt marsh wetland environment with each first mechanical claw gripping force as set value. Each first mechanical claw gripping force corresponds to an independent operation video.

[0056] Step S6: Determine the target mechanical claw gripping force based on the robot's video capture of small arthropods operating in the salt marsh wetland under the gripping force of the multiple first mechanical claws.

[0057] In some embodiments, Figure 3 This is a flowchart illustrating the process of determining the gripping force of a target mechanical claw according to an embodiment of the present invention. The determination of the gripping force includes steps S21 to S24:

[0058] Step S21: Construct a mechanical gripper gripping map. The mechanical gripper gripping map includes multiple first mechanical gripper gripping force nodes and multiple edges between the multiple first mechanical gripper gripping force nodes. The multiple first mechanical gripper gripping force nodes are connected in order of increasing mechanical gripper gripping force. The node features of each first mechanical gripper gripping force node include video of the robot capturing small arthropods in salt marsh wetlands under each first mechanical gripper gripping force. The edge is the force difference of the mechanical gripper gripping force of two first mechanical gripper gripping force nodes.

[0059] A robotic gripper atlas is a type of structured data used to describe the relationships between different gripping forces. In a robotic gripper atlas, each node represents a specific first robotic gripper gripping force. The node features of each first robotic gripper gripping force node include video footage of the robot capturing small arthropods in salt marshes under each first robotic gripper gripping force. The edges between nodes represent the numerical difference in the first robotic gripper gripping forces corresponding to two first robotic gripper gripping force nodes.

[0060] Mechanical gripper gripping patterns can organize originally discrete force test points into an ordered, structured whole that reflects the gradual relationship of force changes.

[0061] Step S22: Process the mechanical gripper gripping map based on the graph neural network to determine the gripping force of multiple second mechanical grippers.

[0062] Graph Neural Networks (GNNs) are deep learning models capable of processing graph data. GNNs capture the topological features of a graph and the attribute features of individual nodes by transmitting and aggregating information between nodes. The input to the GNN is the mechanical gripper's grasping graph, and the output is the gripping force of multiple second mechanical grippers.

[0063] The multiple second mechanical gripper gripping forces are a set of candidate mechanical gripper gripping forces that are more in line with the stable gripping standards and biological protection requirements, selected from multiple first mechanical gripper gripping forces after analyzing and processing the mechanical gripper gripping map through a graph neural network.

[0064] In the robotic gripper atlas, each node corresponds to a first gripper force. Node features include videos of the robot capturing small arthropods in salt marshes at that force. These videos contain core information directly reflecting the capture effect, such as details of the gripping action and the real-time reactions of the organisms. Edges between nodes represent the force differences between adjacent gripping forces. These differences reflect the gradient correlation between different forces and can reflect the trend of force changes on the capture effect, such as the gradual change in the organism's survival status when the force increases slightly. By processing this robotic gripper atlas using a graph neural network, a second robotic gripper force with better overall performance can be selected from both the numerical and effect correlations of the gripping forces. This improves the efficiency and accuracy of selecting target gripping forces, thereby avoiding ineffective force attempts.

[0065] Graph Neural Networks (GNNs) can extract features from video footage of a robot capturing small arthropods in a salt marsh wetland under different gripping forces, based on the node features of each node in the robotic gripper's grasping map. These features are then transformed into computationally readable associated feature codes, while preserving edge gripping force differences and node topological connections. The GNN updates the features of each node in the robotic gripper's grasping map through graph convolution operations. During convolution, each node aggregates the feature information of its neighboring nodes, adjusting the aggregation weights based on edge gripping force differences. Through multiple rounds of graph convolution, the associated feature codes of each node gradually fuse their own capture performance information with the association information of neighboring nodes, forming a feature representation that comprehensively reflects the relative merits of that gripping force within the global robotic gripper grasping map. The GNN then uses fully connected layers to classify and score the updated node features, constructing comprehensive evaluation metrics, including capture performance scores and biodiversity conservation performance scores based on video features. The model can sort all the first mechanical gripper clamping force nodes according to the comprehensive evaluation score, and select the first mechanical gripper clamping force corresponding to the top-ranked nodes as multiple second mechanical gripper clamping forces.

[0066] Step S23: Based on the robot capture operation video of the small arthropods in the salt marsh wetland under different mechanical claw gripping forces, the robot capture operation video of the small arthropods in the salt marsh wetland under multiple first mechanical claw gripping forces, and the multiple second mechanical claw gripping forces, generate a robot capture operation simulation video of the small arthropods in the salt marsh wetland under each second mechanical claw gripping force.

[0067] In some embodiments, a simulation model can be used to generate simulation videos of a robot capturing small arthropods in salt marshes at each second gripping force of the robotic claw. The simulation model is a variational autoencoder. The inputs to the simulation model are the robot capture videos of the small arthropods in salt marshes at different gripping forces of the robotic claws, the robot capture videos of the small arthropods in salt marshes at the gripping forces of the plurality of first robotic claws, and the gripping forces of the plurality of second robotic claws. The output of the simulation model is the robot capture simulation video of the small arthropods in salt marshes at each gripping force of the second robotic claw.

[0068] A Variational Autoencoder (VAE) is a deep learning generative model based on a probabilistic model. It consists of three parts: an encoder, a decoder, and a latent space. The encoder maps the input data to the latent space and generates latent variables with a specific probability distribution. The latent space constrains the variables output by the encoder to follow a predefined probability distribution and enhances the stability and diversity of the generated results. The decoder can reverse-engineer output data with a structure similar to the input data based on the variables in the latent space. By minimizing reconstruction error and using KL divergence regularization, the VAE learns the latent distribution patterns of the input data and can generate new samples that conform to the data characteristics.

[0069] The simulation video of the robot capturing small arthropods in the salt marsh wetland under each gripping force of the second mechanical claw is generated by the simulation model.

[0070] Videos of robot capture operations of small arthropods in salt marshes under different gripping forces of the robotic claws, as well as multiple gripping forces of the first robotic claw, provide the model with a wealth of real-world capture scenario data under varying forces. This data includes potential patterns such as the movement of the robotic claws, the reaction characteristics of the captured small arthropods, and the correlation between capture effectiveness and force. This provides a rich sample base for the model to learn and generate simulation videos. The multiple gripping forces of the second robotic claws clarify the target force values ​​corresponding to the simulation videos and enable the model to generate simulation scenarios specifically for each force, ensuring accurate matching between the simulation videos and the target forces.

[0071] The variational autoencoder (VAE) preprocesses robot capture videos of small arthropods in salt marshes under different gripping forces of the robotic claws, as well as videos of the robot capturing small arthropods in salt marshes under multiple first-claw gripping forces. This preprocessing extracts keyframe features from the videos, including claw posture features, morphological features of the captured target small arthropods, and temporal features of the capture actions. The encoder maps these features to a latent space to generate latent variables following a pre-defined Gaussian distribution. Simultaneously, the encoder learns the distribution patterns of latent variables corresponding to different gripping forces, thus establishing a mapping relationship between gripping forces and latent variables. Based on the learned mapping relationship between gripping forces and latent variables, the model generates latent variables corresponding to each second-claw gripping force in the latent space. These latent variables fully integrate the video feature patterns corresponding to real gripping forces similar to the second-claw gripping force, ensuring the reasonableness of the generated simulation video. The decoder receives these latent variables and gradually reconstructs the video frame sequence through a reverse mapping process. During decoding, the decoder generates simulation video frames with a structure consistent with the real video based on the feature information in the latent variables. For example, the decoder can generate the opening and closing motion and trajectory of the robotic gripper under a given force based on the characteristics of its posture. It can also generate the state changes and morphological responses of an organism under that force based on biological response characteristics. Furthermore, the decoder can concatenate video frames according to temporal characteristics to create a continuous and smooth simulation video. Simultaneously, the model can continuously optimize the generation process through reconstruction error feedback to ensure that the simulation video is highly consistent with the real video in terms of motion logic, biological responses, and capture effects, and accurately reflects the actual capture scenario under the corresponding second robotic gripper force.

[0072] Step S24: Determine the target mechanical claw gripping force based on the robot's simulation video of small arthropods operating in the salt marsh wetland, using the gripping force of each second mechanical claw.

[0073] In some embodiments, a target force determination model can be used to determine the gripping force of the target robotic gripper. The target force determination model is a Transformer model. The input to the target force determination model is a simulation video of the robot capturing the movement of small arthropods in a salt marsh wetland at each of the second robotic gripping forces, and the output of the target force determination model is the gripping force of the target robotic gripper.

[0074] The target gripping force is determined by comprehensively analyzing simulation videos of the robot capturing small arthropods in salt marshes using the target force determination model. This analysis yields the optimal gripping force value that best meets the capture requirements.

[0075] The simulation videos of the robot capturing small arthropods in salt marshes at each second robotic gripper gripping force can fully simulate the entire capture process under the corresponding gripping force, including key information such as the gripper's action effects, the organism's survival status, and shape changes. These simulation videos provide the model with comprehensive performance evaluation data and allow the model to intuitively compare the overall performance of different second robotic gripper gripping forces, thereby accurately identifying the optimal gripping force value.

[0076] The Transformer model's self-attention mechanism endows it with powerful global correlation analysis and precise quantitative evaluation capabilities. This allows the model to deeply analyze simulation videos of robot capture of small arthropods in salt marshes at each second gripping force. The self-attention mechanism can overcome temporal constraints to simultaneously capture key features at different temporal nodes in the simulation videos of robot capture of small arthropods in salt marshes at each second gripping force, and can uncover the intrinsic correlations between temporal nodes without relying on a fixed temporal window, thus quickly locating core information related to gripping effectiveness. By deeply comparing the feature representations of robot capture of small arthropods in salt marshes corresponding to multiple second gripping forces, the Transformer model can quantitatively distinguish the performance differences of each second gripping force in dimensions such as capture stability, organism survival assurance, and morphological damage control. Furthermore, the model can accurately select the value with the best overall performance as the target gripping force.

[0077] In some embodiments, determining the target mechanical gripper gripping force based on the robot's operation simulation video of capturing small arthropods in salt marsh wetlands, according to the gripping force of each of the second mechanical grippers, includes steps S31 to S33:

[0078] Step S31: Based on the robot capture simulation video of small arthropods in salt marsh wetlands with the gripping force of each second mechanical claw, determine the simulated morphological change sequence and the simulated fluctuation sequence of activity amplitude of the small arthropods in salt marsh wetlands corresponding to the gripping force of each second mechanical claw.

[0079] In some embodiments, the Transformer model can be used to determine the simulated morphological change sequence and the simulated activity amplitude fluctuation sequence of small arthropods in salt marshes corresponding to the gripping force of each second mechanical claw.

[0080] The simulated morphological change sequence of small arthropods in salt marshes corresponding to each gripping force of the second mechanical claw is continuous data of the temporal change of the morphology of small arthropods in salt marshes during the capture process under each gripping force of the second mechanical claw, output by the Transformer model.

[0081] The activity amplitude simulation fluctuation sequence is a continuous data set of the overall activity amplitude of small arthropods in salt marsh wetlands fluctuating over time during the capture process, under the gripping force of each second mechanical claw, output by the Transformer model.

[0082] The Transformer model, through its self-attention mechanism, enables in-depth analysis of temporal information in simulation videos of robots capturing small arthropods in salt marshes. It can simultaneously focus on key features in different frames to accurately capture subtle simulated changes in the organism's surface morphology and activity amplitude. The encoder effectively extracts the morphological evolution patterns of the organism's surface during the grasping process, forming a continuous sequence of simulated surface morphological changes. The decoder, based on the encoded features, further mines the fluctuation patterns of the organism's activity amplitude over time, thereby generating a corresponding simulated activity amplitude fluctuation sequence.

[0083] Step S32: Based on the simulated morphological change sequence and activity amplitude fluctuation sequence of small arthropods in salt marsh wetlands corresponding to the gripping force of each second mechanical claw, determine the capture stability score, reversible morphological recovery rate, and morphological protection compliance level under the gripping force of each second mechanical claw.

[0084] In some embodiments, the Transformer model can be used to determine the capture stability score, biomorphic reversibility rate, and morphological protection compliance level under each second mechanical gripper clamping force.

[0085] The capture stability score for each second mechanical claw gripping force is a numerical indicator output by the Transformer model that quantifies the degree of stability of the gripping state of small arthropods in salt marshes under the second mechanical claw gripping force, indicating that they do not fall off.

[0086] The reversible morphological recovery rate is an indicator of the proportion of small arthropods in salt marshes that recover to their initial morphology under the corresponding second mechanical gripping force, as output by the Transformer model.

[0087] The morphological protection compliance level is a grading standard based on whether the morphology of small arthropods in salt marshes meets the requirement of no irreversible damage under the corresponding second mechanical claw gripping force output by the Transformer model.

[0088] The Transformer model can deeply mine key information and intrinsic relationships in simulated morphological change sequences and simulated amplitude fluctuation sequences. For simulated amplitude fluctuation sequences, the model can accurately quantify core features such as fluctuation frequency and amplitude extremes, and then output a capture stability score reflecting the stability of the clamping state through feature aggregation and quantification calculations. For simulated morphological change sequences, the model can extract key parameters such as the amplitude, rate, and recovery trend of morphological changes to calculate the proportion of biological morphology that recovers to its initial state, thereby obtaining the reversible recovery rate of biological morphology. In addition, the model can also compare with the criteria for no irreversible damage to complete the classification of morphological protection compliance levels.

[0089] Step S33: Determine the target mechanical gripper gripping force based on the capture stability score, biological morphology reversible recovery rate, and morphology protection compliance level of each second mechanical gripper gripping force.

[0090] In some embodiments, a deep neural network can be used to determine the gripping force of the target robotic gripper.

[0091] Deep neural networks, through multi-layer nonlinear transformations, can comprehensively model and weight the capture stability score, biomorphic reversible recovery rate, and morphological protection compliance level. The hidden layers of the deep neural network can automatically learn the complex relationships between these three indicators. By continuously optimizing parameters through backpropagation, the model can accurately quantify the influence weight of different indicators in the final decision. For example, while ensuring that the morphological protection compliance level meets the requirements, the weight of the capture stability score can be prioritized. By comprehensively evaluating and ranking multiple indicators corresponding to the gripping force of each second manipulator, the deep neural network can select the value with the best overall performance as the target manipulator gripping force. This achieves a precise mapping from multi-dimensional evaluation of the capture stability score, biomorphic reversible recovery rate, and morphological protection compliance level under each second manipulator gripping force to a single optimal solution.

[0092] Step S7: Based on the gripping force of the target mechanical claw, the robot is subjected to PID servo drive control to complete the capture operation of small arthropods in the salt marsh wetland.

[0093] Once the target gripping force of the mechanical claw is determined, PID servo drive control is executed on the robot based on the target gripping force, so that the robot's mechanical claw accurately outputs the target gripping force, and completes the stable capture of small arthropods in the salt marsh wetland environment.

[0094] Based on the same inventive concept Figure 4 This is a schematic diagram of a PID-based servo drive control system provided in an embodiment of the present invention. The PID-based servo drive control system includes:

[0095] The first acquisition module 41 is used to acquire robot capture operation videos under different mechanical claw gripping forces of small arthropods in salt marshes;

[0096] The task information determination module 42 is used to determine the task information for each mechanical claw gripping force based on the robot capture operation video under different mechanical claw gripping forces of small arthropods in the salt marsh wetland using the capture performance determination model.

[0097] The clamping force range determination module 43 is used to determine the lower limit clamping force of the anti-detachment mechanical claw and the upper limit clamping force of the life-saving mechanical claw based on the operation information under the clamping force of each mechanical claw.

[0098] The first clamping force determination module 44 is used to determine multiple first mechanical claw clamping forces based on the clamping force of the anti-detachment lower limit mechanical claw, the clamping force of the life-protecting upper limit mechanical claw, and the operation information under the clamping force of each mechanical claw.

[0099] The second acquisition module 45 is used to acquire multiple videos of the robot capturing small arthropods in salt marshes under the gripping force of the first mechanical claw.

[0100] The target gripping force determination module 46 is used to determine the target gripping force of the mechanical claw based on the robot's video of capturing small arthropods in salt marsh wetlands under the gripping force of the plurality of first mechanical claws.

[0101] The control execution module 47 is used to perform PID servo drive control on the robot based on the gripping force of the target mechanical claw, so as to complete the capture operation of small arthropods in the salt marsh wetland.

[0102] It should be noted that, in order to simplify the descriptions disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments of this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0103] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A servo drive control method based on PID, characterized in that, include Videos of robot capture and operation under different gripping forces of mechanical claws on small arthropods in salt marshes were obtained. Based on the robot capture operation video of small arthropods in the salt marsh wetland under different mechanical claw gripping forces, the capture performance determination model is used to determine the operation information under each mechanical claw gripping force. The lower limit of anti-detachment mechanical claw gripping force and the upper limit of life-saving mechanical claw gripping force are determined based on the operational information under the gripping force of each mechanical claw. Based on the lower limit gripping force of the anti-detachment mechanical claw, the upper limit gripping force of the life-protection mechanical claw, and the operational information under each mechanical claw gripping force, multiple first mechanical claw gripping forces are determined. Acquire multiple videos of the robot capturing small arthropods in salt marshes under the gripping force of the first mechanical claw; The target mechanical gripper's gripping force is determined based on video footage of the robot capturing small arthropods in salt marshes under the gripping forces of the multiple first mechanical grippers. This determination of the target mechanical gripper's gripping force based on the video footage of the robot capturing small arthropods in salt marshes under the gripping forces of the multiple first mechanical grippers includes: A mechanical gripper gripping atlas is constructed, which includes multiple first mechanical gripper gripping force nodes and multiple edges between the multiple first mechanical gripper gripping force nodes. The multiple first mechanical gripper gripping force nodes are connected in order of increasing mechanical gripper gripping force. The node features of each first mechanical gripper gripping force node include video of the robot capturing small arthropods in salt marsh wetlands under each first mechanical gripper gripping force. The edge is the force difference of mechanical gripper gripping force between two first mechanical gripper gripping force nodes. The gripping force of multiple second mechanical claws is determined by processing the gripping map of the mechanical claw based on a graph neural network. Based on the robot capture operation video under different mechanical claw gripping forces of the small arthropods in the salt marsh wetland, the robot capture operation video under multiple first mechanical claw gripping forces of the small arthropods in the salt marsh wetland, and the multiple second mechanical claw gripping forces, a robot capture operation simulation video of the small arthropods in the salt marsh wetland is generated for each second mechanical claw gripping force. The target mechanical gripper gripping force is determined by capturing simulation videos of small arthropods operating in salt marshes based on the gripping force of each second mechanical gripper. Based on the gripping force of the target mechanical claw, the robot is controlled by a PID servo drive to complete the capture of small arthropods in the salt marsh wetland.

2. The servo drive control method based on PID as described in claim 1, characterized in that, The operational information under the gripping force of each mechanical claw includes the survival status information and shape change information of the target organism.

3. The servo drive control method based on PID as described in claim 1, characterized in that, The input to the graph neural network is the mechanical gripper pattern, and the output of the graph neural network is the gripping force of multiple second mechanical grippers.

4. A servo drive control system based on PID, characterized in that, include: The first acquisition module is used to acquire robot capture operation videos under different mechanical claw gripping forces of small arthropods in salt marshes; The task information determination module is used to determine the task information for each mechanical claw gripping force based on the robot's capture operation video under different mechanical claw gripping forces of small arthropods in the salt marsh wetland, using a capture performance determination model. The clamping force range determination module is used to determine the lower limit clamping force of the anti-detachment mechanical claw and the upper limit clamping force of the life-saving mechanical claw based on the operation information under the clamping force of each mechanical claw. The first clamping force determination module is used to determine multiple first mechanical claw clamping forces based on the clamping force of the anti-detachment lower limit mechanical claw, the clamping force of the life-protecting upper limit mechanical claw, and the operation information under each mechanical claw clamping force. The second acquisition module is used to acquire multiple videos of the robot capturing small arthropods in salt marshes under the gripping force of the first mechanical claw. The target gripping force determination module is used to determine the target gripping force of the robotic claw based on video footage of the robot capturing small arthropods in salt marsh wetlands under the gripping forces of the plurality of first robotic claws. The target gripping force determination module is also used for: A mechanical gripper gripping atlas is constructed, which includes multiple first mechanical gripper gripping force nodes and multiple edges between the multiple first mechanical gripper gripping force nodes. The multiple first mechanical gripper gripping force nodes are connected in order of increasing mechanical gripper gripping force. The node features of each first mechanical gripper gripping force node include video of the robot capturing small arthropods in salt marsh wetlands under each first mechanical gripper gripping force. The edge is the force difference of mechanical gripper gripping force between two first mechanical gripper gripping force nodes. The gripping force of multiple second mechanical claws is determined by processing the gripping map of the mechanical claw based on a graph neural network. Based on the robot capture operation video under different mechanical claw gripping forces of the small arthropods in the salt marsh wetland, the robot capture operation video under multiple first mechanical claw gripping forces of the small arthropods in the salt marsh wetland, and the multiple second mechanical claw gripping forces, a robot capture operation simulation video of the small arthropods in the salt marsh wetland is generated for each second mechanical claw gripping force. The target mechanical gripper gripping force is determined by capturing simulation videos of small arthropods operating in salt marshes based on the gripping force of each second mechanical gripper. The control execution module is used to perform PID servo drive control on the robot based on the gripping force of the target mechanical claw, so as to complete the capture operation of small arthropods in the salt marsh wetland.

5. The PID-based servo drive control system as described in claim 4, characterized in that, The operational information under the gripping force of each mechanical claw includes the survival status information and shape change information of the target organism.

6. The PID-based servo drive control system as described in claim 4, characterized in that, The input to the graph neural network is the mechanical gripper pattern, and the output of the graph neural network is the gripping force of multiple second mechanical grippers.

7. An electronic device, characterized in that, include: processor; Memory; And a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the PID-based servo drive control method as described in any one of claims 1 to 3.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the PID-based servo drive control method as described in any one of claims 1 to 3.

Images