Autonomous control learning system and method of electromagnetic drive surgical robot

Abstract

The invention particularly relates to an autonomous control learning system and method of an electromagnetic drive surgical robot, and the system comprises a virtual simulation module and an electromagnetic drive system, the electromagnetic drive system comprises an electromagnetic coil module, a hardware drive module and an artificial intelligence calculation module, the virtual simulation module is used for simulating behaviors of the electromagnetic drive surgical robot in a preset virtual surgical environment to obtain first modal data; the electromagnetic driving system is used for receiving the first modal data and obtaining second modal data based on the data through actual operation of the electromagnetic coil module and the hardware driving module; the artificial intelligence calculation module generates an optimal control signal of the electromagnetic coil module through a preset deep learning model and a preset diffusion model based on the two modal data, and drives the electromagnetic drive surgical robot to autonomously complete tasks. Therefore, the problems of poor adaptability, insufficient flexibility and the like of an electromagnetic drive surgical robot in the related technology are solved, and the operation flexibility is improved.

Description

Technical Field

[0001] This application relates to the field of robot technology, and particularly to an autonomous control learning system and method for an electromagnetic drive surgical robot. Background Art

[0002] Electromagnetic drive surgical robots are gradually becoming a research hotspot in the field of medical robots due to their advantages of high precision and non-contact operation. In surgical scenarios such as puncture, endoscopy, and intervention, they show extensive application potential. In recent years, researchers have proposed various training schemes for novices in virtual simulation, and designed and explored autonomous control strategies based on imitation learning and reinforcement learning specifically for electromagnetic drive surgical robots. At the same time, in the physical experimental environment, the research usually combines common surgical instruments such as needles, capsules, and guide wires to reveal more detailed motion characteristics and complex dynamics than simulation.

[0003] In the related art, the working process of an electromagnetic drive surgical robot can generally be divided into a target stage, a planning stage, and an execution stage. The target stage aims to diagnose and identify surgical targets, such as diseased tissues or biopsy sites; the planning stage is responsible for determining the motion trajectory of the micro-surgical instrument to ensure the efficiency and controllability of the operation; the execution stage then converts the planned trajectory into continuous actions of the robot, which can be specifically realized through pre-performed finite element simulation or real-time calculation based on an online solver. In addition, the step-by-step strategies in the related art usually adopt a multi-step framework, for example, first estimate the environmental condition parameters, then determine the motion trajectory, and finally solve the mapping from the electromagnetic field to current control.

[0004] However, the adaptability of the methods in the related art is limited, and the flexibility and safety of electromagnetic field planning still need to be improved. It is difficult to achieve optimal planning, which further limits the flexibility and real-time response ability of the system and urgently needs to be solved. Summary of the Invention

[0005] This application provides an autonomous control learning system and method for an electromagnetic drive surgical robot to solve the problems of poor adaptability and insufficient flexibility of the electromagnetic drive surgical robot in the environment in the related art, and improve the operation flexibility.

[0006] The first aspect embodiment of this application provides an autonomous control learning system for an electromagnetic drive surgical robot, including: a virtual simulation module and an electromagnetic drive system. The electromagnetic drive system includes an electromagnetic coil module, a hardware drive module, and an artificial intelligence calculation module, where,

[0007] The virtual simulation module is used to simulate the behavior of the electromagnetic drive surgical robot in a preset virtual surgical environment to obtain first modal data;

[0008] The electromagnetic drive system is used to receive the first modal data, and based on the first modal data, through the actual operation of the electromagnetic coil module and the hardware drive module, obtain the second modal data;

[0009] The artificial intelligence computing module generates an optimal control signal for the electromagnetic coil module based on the first modal data and the second modal data through a preset deep learning model and a preset diffusion model, and drives the electromagnetic drive surgical robot to autonomously complete tasks based on the optimal control signal.

[0010] Optionally, the virtual simulation module includes:

[0011] A reconstruction unit, configured to obtain medical image data, and based on the medical image data, perform three-dimensional reconstruction on a human organ or tissue structure to obtain a three-dimensional reconstruction result;

[0012] A first generation unit, configured to generate the preset virtual surgical environment based on the three-dimensional reconstruction result, simulate the motion trajectory, electromagnetic field distribution, and hydrodynamic parameters of the electromagnetic drive surgical robot in the preset virtual surgical environment, and obtain the first modal data according to the simulation result.

[0013] Optionally, the second modal data includes at least one of the current of the electromagnetic coil, the magnetic field strength, the actual position of the surgical instrument, and the attitude of the surgical instrument.

[0014] Optionally, the electromagnetic drive system further includes a vision unit module, wherein,

[0015] The hardware drive module determines the current of the electromagnetic coil according to the optimal control signal;

[0016] The electromagnetic coil module includes a plurality of electromagnetic coils, and the plurality of electromagnetic coils generate the magnetic field strength based on the current of the electromagnetic coil;

[0017] The vision unit module is configured to collect the actual position of the surgical instrument of the electromagnetic drive surgical robot and the attitude of the surgical instrument.

[0018] Optionally, the electromagnetic drive system further includes:

[0019] A wireless communication module, configured to send the second modal data to the virtual simulation module, so that the virtual simulation module adjusts the simulation parameters of the preset virtual surgical environment according to the second modal data.

[0020] Optionally, the artificial intelligence computing module includes:

[0021] A receiving unit, configured to receive the first modal data and the second modal data;

[0022] A processing unit, configured to extract features and perform feature characterization on the first-modal data and the second-modal data by using the preset deep learning model, and based on the processing result, plan an electromagnetic field by using the preset diffusion model to obtain the optimal control signal.

[0023] An embodiment of the second aspect of the present application provides an autonomous control learning method for an electromagnetic drive surgical robot, which adopts the autonomous control learning system of the electromagnetic drive surgical robot in the embodiment of the first aspect, and includes the following steps:

[0024] Simulate the behavior of the electromagnetic drive surgical robot in a preset virtual surgical environment through the virtual simulation module to obtain first-modal data;

[0025] Receive the first-modal data through the electromagnetic drive system, and based on the first-modal data, obtain second-modal data through the actual operation of the electromagnetic coil module and the hardware drive module;

[0026] Generate the optimal control signal of the electromagnetic coil module based on the first-modal data and the second-modal data through the artificial intelligence computing module by using a preset deep learning model and a preset diffusion model, and drive the electromagnetic drive surgical robot to autonomously complete tasks based on the optimal control signal.

[0027] Optionally, the step of simulating the behavior of the electromagnetic drive surgical robot in a preset virtual surgical environment through the virtual simulation module to obtain first-modal data includes:

[0028] Obtain medical image data through a reconstruction unit, and based on the medical image data, perform three-dimensional reconstruction on a human organ or tissue structure to obtain a three-dimensional reconstruction result;

[0029] Generate the preset virtual surgical environment based on the three-dimensional reconstruction result through a first generation unit, simulate the motion trajectory, electromagnetic field distribution, and hydrodynamic parameters of the electromagnetic drive surgical robot in the preset virtual surgical environment, and obtain the first-modal data according to the simulation result.

[0030] Optionally, the second-modal data includes at least one of the current of the electromagnetic coil, the magnetic field strength, the actual position of the surgical instrument, and the attitude of the surgical instrument.

[0031] Optionally, the electromagnetic drive system further includes:

[0032] Determine the current of the electromagnetic coil according to the optimal control signal through the hardware drive module;

[0033] Through a plurality of electromagnetic coils of the electromagnetic coil module, the plurality of electromagnetic coils generate the magnetic field intensity based on the current of the electromagnetic coils;

[0034] The vision unit module is used to collect the actual position of the surgical instrument of the electromagnetic drive surgical robot and the posture of the surgical instrument.

[0035] Optionally, the electromagnetic drive system further includes:

[0036] The second modality data is sent to the virtual simulation module through the wireless communication module, so that the virtual simulation module adjusts the simulation parameters of the preset virtual surgical environment according to the second modality data.

[0037] Optionally, based on the first modality data and the second modality data, the artificial intelligence calculation module generates an optimal control signal for the electromagnetic coil module through a preset deep learning model and a preset diffusion model, and drives the electromagnetic drive surgical robot to autonomously complete tasks, including:

[0038] Receiving the first modality data and the second modality data through a receiving unit;

[0039] The processing unit uses the preset deep learning model to perform feature extraction and feature characterization on the first modality data and the second modality data, and based on the processing results, plans the electromagnetic field through the preset diffusion model to obtain the optimal control signal.

[0040] Thus, in the embodiment of the present application, the virtual simulation module simulates the behavior of the electromagnetic drive surgical robot in a preset virtual surgical environment to obtain the first modality data. The electromagnetic drive system receives the first modality data, and based on this data, through the actual operation of the electromagnetic coil module and the hardware drive module, obtains the second modality data. The artificial intelligence calculation module generates an optimal control signal for the electromagnetic coil module based on the first modality data and through a preset deep learning model and a preset diffusion model, and drives the electromagnetic drive surgical robot to autonomously complete tasks. Thus, the problems of poor adaptability and insufficient flexibility of the electromagnetic drive surgical robot in the environment in the related art are solved, and the operation flexibility is improved.

[0041] Additional aspects and advantages of the present application will be given in part in the following description, become apparent in part from the following description, or be understood through the practice of the present application. Description of the Drawings

[0042] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description of the embodiments in conjunction with the drawings, where:

[0043] Figure 1Schematic block diagram of an autonomous control learning system for an electromagnetic drive surgical robot provided according to an embodiment of the present application;

[0044] FIG. 2 is a schematic diagram of an electromagnetic coil array and simulation of an autonomous control learning system for an electromagnetic drive surgical robot provided according to an embodiment of the present application;

[0045] Figure 3 Schematic diagram of the process of generating and adding noise to an electromagnetic field by a diffusion model of an autonomous control learning system for an electromagnetic drive surgical robot provided according to an embodiment of the present application;

[0046] Figure 4 Schematic diagram of a virtual simulation environment of an autonomous control learning system for an electromagnetic drive surgical robot provided according to an embodiment of the present application;

[0047] Figure 5 Schematic diagram of a feasible algorithm framework of an artificial intelligence computing module of an autonomous control learning system for an electromagnetic drive surgical robot provided according to an embodiment of the present application;

[0048] Figure 6 Schematic diagram of the structure of a deep learning model of an autonomous control learning system for an electromagnetic drive surgical robot provided according to an embodiment of the present application;

[0049] Figure 7 Flowchart of an autonomous control learning method for an electromagnetic drive surgical robot provided according to an embodiment of the present application. Detailed implementation manners

[0050] The embodiments of the present application will be described in detail below. The examples of the embodiments are shown in the drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are intended to explain the present application, but should not be construed as a limitation to the present application.

[0051] The autonomous control learning system and method of an electromagnetic drive surgical robot according to an embodiment of the present application will be described below with reference to the accompanying drawings. In view of the problems of poor adaptability and insufficient flexibility of the electromagnetic drive surgical robot in the environment in the related art mentioned in the above background art, the present application provides an autonomous control learning system for an electromagnetic drive surgical robot, which includes: a virtual simulation module and an electromagnetic drive system. The electromagnetic drive system includes an electromagnetic coil module, a hardware drive module, and an artificial intelligence calculation module. The virtual simulation module is used to simulate the behavior of the electromagnetic drive surgical robot in a preset virtual surgical environment to obtain first modal data; the electromagnetic drive system is used to receive the first modal data and, based on this data, obtain second modal data through the actual operation of the electromagnetic coil module and the hardware drive module; the artificial intelligence calculation module generates an optimal control signal for the electromagnetic coil module based on these two modal data through a preset deep learning model and a preset diffusion model, and drives the electromagnetic drive surgical robot to autonomously complete tasks. Thus, the problems of poor adaptability and insufficient flexibility of the electromagnetic drive surgical robot in the environment in the related art are solved, and the operation flexibility is improved.

[0052] Specifically, Figure 1 The block diagram of an autonomous control learning system for an electromagnetic drive surgical robot provided by an embodiment of the present application.

[0053] As Figure 1 shown, the autonomous control learning system 10 of the electromagnetic drive surgical robot includes: a virtual simulation module 100 and an electromagnetic drive system 200. The electromagnetic drive system 200 includes an electromagnetic coil module 300, a hardware drive module 400, and an artificial intelligence calculation module 500.

[0054] Among them, the virtual simulation module 100 is used to simulate the behavior of the electromagnetic drive surgical robot in a preset virtual surgical environment to obtain first modal data;

[0055] The electromagnetic drive system 200 is used to receive the first modal data and, based on the first modal data, obtain second modal data through the actual operation of the electromagnetic coil module 300 and the hardware drive module 400;

[0056] The artificial intelligence calculation module 500 generates an optimal control signal for the electromagnetic coil module based on the first modal data and the second modal data through a preset deep learning model and a preset diffusion model, and drives the electromagnetic drive surgical robot to autonomously complete tasks based on the optimal control signal.

[0057] Specifically, by synchronously acquiring teaching data on the virtual simulation module 100 and the electromagnetic drive system 200 for joint training, using a vision model and a regression model for task representation, and implementing an optimized planning of the electromagnetic field through a diffusion model, it is possible to directly predict the optimal control signal accurate to each fixed electromagnetic coil from virtual and actual video signals, thereby controlling the electromagnetic drive surgical robot to autonomously complete tasks. Thus, the virtual simulation module 100 provides high-quality simulation data for the embodiments of the present application, helping to construct a virtual surgical environment and simulate the behavior of the surgical robot; the electromagnetic drive system 200 operates on an actual hardware platform, generates actual observation data, and aligns and fuses it with the virtual simulation data; the artificial intelligence computing module 500 generates the optimal control signal through a deep learning model and a diffusion model, driving the surgical robot to autonomously complete tasks; each module works collaboratively to achieve a complete process from virtual simulation to actual hardware operation, ensuring the high precision, real-time performance, and intelligence of the embodiments of the present application.

[0058] Optionally, in some embodiments, the virtual simulation module 100 includes: a reconstruction unit, configured to acquire medical image data and perform three-dimensional reconstruction on a human organ or tissue structure based on the medical image data to obtain a three-dimensional reconstruction result; a first generation unit, configured to generate a preset virtual surgical environment based on the three-dimensional reconstruction result, simulate the motion trajectory, electromagnetic field distribution, and hydrodynamic parameters of the electromagnetic drive surgical robot in the preset virtual surgical environment, and obtain first modal data according to the simulation result.

[0059] The first modal data refers to the data simulated in the preset virtual surgical environment generated based on the three-dimensional reconstruction result.

[0060] It can be understood that based on the method of equal-proportion modeling, three-dimensional reconstruction of a human organ or tissue structure is performed through clinically acquired medical image data (including but not limited to CT, MRI, B-ultrasound, etc.) data, and fine-tuning and rendering are performed on this basis to obtain a three-dimensional reconstruction result, completing the simulation and emulation of the human organ or tissue structure and generating a preset virtual surgical environment. The preset virtual surgical environment includes but is not limited to puncture surgery, ablation surgery, endoscopic surgery, interventional surgery, etc.; in the virtual surgical environment, the first generation unit can simulate the motion trajectory of the electromagnetic drive surgical robot, including but not limited to the movement, turning of surgical instruments, and interaction with other tissues or organs, etc., and obtain first modal data according to the simulation result.

[0061] Optionally, in some embodiments, the second modal data includes at least one of the current of the electromagnetic coil, the magnetic field strength, the actual position of the surgical instrument, and the attitude of the surgical instrument.

[0062] Among them, the second-modal data are real-time monitoring and control data directly related to the operation of the electromagnetic drive surgical robot; the current of the electromagnetic coil is the current intensity passing through the electromagnetic coil to generate the required magnetic field; the magnetic field intensity is the strength of the magnetic field generated by the electromagnetic coil at a specific position; the actual position of the surgical instrument is the specific coordinate position of the surgical instrument in the three-dimensional space during the surgical process; the attitude of the surgical instrument is its direction or angle relative to a certain reference system, usually including three degrees of freedom: pitch, yaw, and roll.

[0063] It can be understood that during the actual surgical process, the second-modal data can be collected in real time and displayed to the doctor through a visualization interface, providing immediate feedback to help the doctor adjust the operation strategy in a timely manner; based on the second-modal data, the motion trajectory and operation process of the surgical robot can be further optimized to reduce the surgical time and risk; in the virtual surgical environment, using the second-modal data, a highly realistic simulation scenario can be created for the doctor to perform preoperative practice.

[0064] Optionally, in some embodiments, the artificial intelligence computing module 500 includes: a receiving unit for receiving the first-modal data and the second-modal data; a processing unit for extracting and characterizing the features of the first-modal data and the second-modal data by using a preset deep learning model, and based on the processing results, planning the electromagnetic field through a preset diffusion model to obtain an optimal control signal.

[0065] It can be understood that through the combination of the deep learning model and the diffusion model, the complex electromagnetic field can be accurately planned to ensure that the surgical instrument moves along the expected trajectory in the virtual environment, improving the accuracy of the surgical operation; the receiving unit continuously receives the first-modal and second-modal data, and the processing unit analyzes and generates a new optimal control signal in real time, realizing closed-loop control, and can dynamically adjust the electromagnetic field distribution according to the real-time feedback to adapt to the changing surgical environment; the diffusion model not only considers the distribution of the electromagnetic field, but also takes into account the optimization of the surgical path and the minimization of energy consumption. By optimizing the path planning, unnecessary energy consumption is reduced and the surgical time is shortened; the data and analysis results generated by the processing unit can provide important decision-making support for the doctor, and the collected first-modal and second-modal data can not only be used for real-time monitoring, but also for detailed analysis after the operation to discover potential problems and improvement points, and further optimize future surgical plans and technical means.

[0066] Furthermore, the artificial intelligence computing module 500 not only processes the first-modal data and the second-modal data, but also further optimizes the working trajectory and control strategy of the surgical robot through various machine learning methods, human-computer interaction data, and content based on large language models. The operation results obtained by sampling, dimensionality reduction, and characterization of the working trajectory of the electromagnetic drive surgical robot through machine learning methods; human-computer interaction data recorded using natural language and multimedia; interaction information of gesture and speech recognition; and electromagnetic drive-related content based on large language models, as well as the weighted combination results of the above methods.

[0067] Thus, the embodiment of the present application first aligns the multi-modal observation data of the physical environment and the virtual environment, and fuses them into hybrid features in the feature space. These features serve as conditional constraints to promote the subsequent electromagnetic field generative planning. By fusing the multi-modal observation data in the virtual environment and the physical environment, while performing tasks in the physical environment, the human-computer interaction in the virtual environment is adjusted, rendered, and previewed in real time. The generative planning of the electromagnetic field is completed by a diffusion model, which converts the noise vector field into a specific electromagnetic field. Through direct planning by the diffusion model and outputting the optimal parameters for controlling the electromagnetic field, it avoids the multi-step processes such as dynamic modeling, motion planning, and field planning in the related art. The generated electromagnetic field is applied to the physical scene to manipulate the micro-surgical instruments, and the generated changes are synchronously transmitted back to the virtual scene for rendering to ensure real-time feedback and precise control. The embodiment of the present application also combines the gesture and speech recognition functions to support multi-modal human-computer interaction, and realizes the complete process from task pre-planning to autonomous control learning through a large language model, which is applicable to various types of systems such as magnetic needles, capsules, and guide wires, showing broad adaptability and application potential.

[0068] Optionally, in some embodiments, the electromagnetic drive system 200 further includes a visual unit module. Among them, the hardware drive module 400 determines the current of the electromagnetic coil according to the optimal control signal; the electromagnetic coil module 300 includes a plurality of electromagnetic coils, and the plurality of electromagnetic coils generate magnetic field intensity based on the current of the electromagnetic coil; the visual unit module is used to collect the actual position of the surgical instrument of the electromagnetic drive surgical robot and the posture of the surgical instrument.

[0069] It can be understood that the visual unit module collects the position and posture data of the surgical instrument in real time, providing real-time feedback for the embodiment of the present application; the hardware drive module 400 adjusts the current output of the electromagnetic coil according to the optimal control signal to ensure the precise control of the magnetic field intensity and the real-time performance and response speed of the embodiment of the present application; the electromagnetic coil module 300 generates an accurate magnetic field based on the current signal to control the movement of the surgical instrument, realizing non-contact operation and improving the safety and accuracy of the surgery.

[0070] Optionally, in some embodiments, the electromagnetic drive system 200 further includes: a wireless communication module configured to send the second modality data to the virtual simulation module 100, so that the virtual simulation module 100 adjusts the simulation parameters of the preset virtual surgical environment according to the second modality data.

[0071] It can be understood that the wireless communication module transmits the second modality data in real time, and the virtual simulation module 100 can dynamically adjust the following simulation parameters according to the actual operation conditions, including the electromagnetic field distribution, that is, update the distribution map of the electromagnetic field in the virtual environment according to the current and magnetic field intensity data of the electromagnetic coil; the position and attitude of the surgical instrument, that is, update the position and attitude of the surgical instrument in the virtual environment in real time to ensure that its movement trajectory is consistent with the actual situation; hydrodynamic parameters such as blood flow velocity, etc., to simulate the impact of the surgery on the physiological state. The embodiments of the present application display the latest simulation results to the doctor through a graphical user interface (GUI) or other visualization tools, providing intuitive operation guidance; allowing the doctor to make interactive modifications in the virtual environment, such as adjusting the surgical path or testing different control strategies to find the best solution.

[0072] The autonomous control learning system of the electromagnetic drive surgical robot will be described in detail below in combination with the autonomous control learning system of the electromagnetic drive surgical robot in a specific embodiment of the present application.

[0073] For a micro-surgical instrument, assuming that its magnetized volume is cylindrical, with a height of h' and a radius of r', the magnitude of the magnetic moment is:

[0074] |μ| = Mπh'r' 2

[0075] where M is the magnetization intensity, the direction of μ is aligned with the long axis of the cylinder, and the unit magnetic force can be expressed as:

[0076]

[0077] Integrating it gives the total electromagnetic force as

[0078]

[0079] where is the integration region corresponding to the magnetized volume of the handheld interactive instrument. Further, if the electromagnetic field at any point in space is denoted as B, and the electromagnetic field distribution in the entire space is denoted as then the relationship between the two is expressed as:

[0080]

[0081] As shown in Figure 2, (a) of Figure 2 is a schematic diagram of an electromagnetic coil array of an autonomous control learning system for an electromagnetic drive surgical robot provided according to an embodiment of the present application; (b) of Figure 2 is a simulation schematic diagram of an electromagnetic coil array of an autonomous control learning system for an electromagnetic drive surgical robot provided according to an embodiment of the present application. Specifically, S is a vector from an arbitrary point in space to the geometric center of the electromagnetic coil, L is a vector from the geometric center of the electromagnetic coil to the origin of the world coordinate system, and I is the current scalar of each electromagnetic coil.

[0082] Establish a learnable electromagnetic field planning strategy, denoted as π Planning , and the form of this strategy is as follows:

[0083]

[0084] Among them, is the electromagnetic field at the i-th step in the teaching trajectory, is the transformation sequence of the electromagnetic fields from the 0-th step to the (i - 1)-th step. Further, π Planning can be decoupled into two parts, namely π Pose and π Current , and the specific forms are as follows:

[0085]

[0086] Because L is the spatial pose relationship between multiple electromagnetic coils in the array and can potentially be accurately obtained from visual information, it is further expressed as:

[0087]

[0088] Among them, π Encoder refers to the backbone visual neural network used to represent encoding, and respectively represent the observation sequences obtained from the physical scene and the virtual scene. Combining the above formulas, we can get:

[0089]

[0090] is equivalent to:

[0091]

[0092] Among them, π Θ is the generation model for electromagnetic field planning, Θ is the specific parameter of this model, and we will introduce in detail how to implement it in combination with the diffusion model in the following. So far, we review the initial electromagnetic force calculation formula and can get:

[0093]

[0094] The following will introduce π in detail Θ 's design; this model uses a diffusion model to achieve electromagnetic field planning through conditional generation. As Figure 3 shown, the desired diffusion model is expressed as:

[0095]

[0096] In this equation, is the electromagnetic field and its distribution in the teaching data, while is the latent variable with the same dimension as . The symbol Θ is the parameter of the diffusion model. The joint distribution is called the inverse process and can be described as a Markov chain:

[0097]

[0098] where,

[0099]

[0100] and represents a standard Gaussian distribution. On the other hand, the forward process is called the diffusion process and is expressed as This process is constructed as a Markov chain, and under the guidance of the variance schedule β1,…,β T , the Gaussian noise is gradually mixed into the electromagnetic field in the teaching data, specifically expressed as:

[0101]

[0102] where,

[0103]

[0104] Let and Then we can express the forward process as:

[0105]

[0106] The above formula shows that given , the distribution of can be directly determined without going through intermediate steps. If the variance is restricted to a time-dependent constant, then given and , the posterior distribution of can be derived as follows:

[0107]

[0108] where,

[0109]

[0110] For electromagnetic field planning, it can be described as a denoising model f Θ , to

[0111]

[0112] and the noise vector field as the input

[0113]

[0114] and finally plan and output For training f Θ The objective function designed is expressed as:

[0115]

[0116] Combining the above, the inverse process of the conditional constraint can be further described as:

[0117]

[0118] Among them,

[0119]

[0120] Once the denoising model f Θ is obtained, according to and estimate ∈, the diffusion process described above can be rearranged to obtain:

[0121]

[0122] Specifically, the expression for each iteration is:

[0123]

[0124] According to the above, the training of the model needs to combine the teaching data in the virtual environment and the actual environment, Figure 4Schematic diagram of the virtual simulation environment of an autonomous control learning system for an electromagnetic drive surgical robot provided by an embodiment of the present application. Taking virtual interventional surgery as an example, it includes two core components: a blood vessel model and a guide wire model. The virtual blood vessel is a proportional model based on the human aorta, and the blood vessel modeling is completed based on medical CT images. In terms of guide wire simulation, a continuum skeletal model is adopted, and the object posture is adjusted through rotation constraints and joint transformation data to achieve accurate simulation of object deformation. In addition, the physical properties such as elasticity and resistance are successfully replicated by wrapping the continuum model with a mesh body for skinning. The skinning technique binds the mesh vertices to the bones, enabling each vertex to respond according to the movement of adjacent bones and their influence weights. Usually, each vertex is affected by less than four bones, and the translation or rotation transformation of the bones will cause corresponding deformation of the mesh body.

[0125] Thus, as Figure 5 shown, Figure 5 Schematic diagram of a feasible algorithm framework for the artificial intelligence calculation module of an autonomous control learning system for an electromagnetic drive surgical robot proposed in the present application, which can autonomously learn and precisely regulate electromagnetic force. This framework integrates the observation data in virtual surgery and the interaction data in the actual system into a stacked input and processes it through a unified encoder. The generated high-dimensional feature vector is reduced in dimension by principal component analysis, spliced and fused with the trajectory data, and after normalization, it is input into the diffusion model as the generation condition for electromagnetic field planning, and finally directly output and converted into coil current to render the electromagnetic field. In the network training stage, the model is supervised by manual annotation to improve its prediction accuracy and stability. Specifically, as Figure 6 shown, the embodiment of the present application can be implemented by using a deep learning model of splicing fusion or attention fusion.

[0126] According to the autonomous control learning system for an electromagnetic drive surgical robot proposed in the embodiment of the present application, in the embodiment of the present application, the virtual simulation module simulates the behavior of the electromagnetic drive surgical robot in a preset virtual surgical environment to obtain first-modal data. The electromagnetic drive system receives the first-modal data and, based on this data, through the actual operation of the electromagnetic coil module and the hardware drive module, obtains second-modal data; the artificial intelligence calculation module, based on the first-modal data and, generates the optimal control signal of the electromagnetic coil module through a preset deep learning model and a preset diffusion model, and drives the electromagnetic drive surgical robot to autonomously complete the task. Thus, the problems of poor adaptability and insufficient flexibility of the electromagnetic drive surgical robot in the environment in the related technology are solved, and the operation flexibility is improved.

[0127] Next, refer to the drawings to describe the autonomous control learning method for an electromagnetic drive surgical robot proposed in the embodiment of the present application.

[0128] In this embodiment, the autonomous control learning method of the electromagnetic drive surgical robot adopts Figure 1 the autonomous control learning system of the electromagnetic drive surgical robot shown in the embodiment.

[0129] As Figure 7 shown, the autonomous control learning method of the electromagnetic drive surgical robot includes the following steps:

[0130] In step S701, the virtual simulation module simulates the behavior of the electromagnetic drive surgical robot in a preset virtual surgical environment to obtain first modal data.

[0131] In step S702, the electromagnetic drive system receives the first modal data, and based on the first modal data, through the actual operation of the electromagnetic coil module and the hardware drive module, second modal data is obtained.

[0132] In step S703, the artificial intelligence calculation module generates an optimal control signal for the electromagnetic coil module based on the first modal data and the second modal data through a preset deep learning model and a preset diffusion model, and drives the electromagnetic drive surgical robot to autonomously complete tasks based on the optimal control signal.

[0133] Optionally, in some embodiments, simulating the behavior of the electromagnetic drive surgical robot in a preset virtual surgical environment through the virtual simulation module to obtain first modal data includes: obtaining medical image data through a reconstruction unit, and based on the medical image data, performing three-dimensional reconstruction on a human organ or tissue structure to obtain a three-dimensional reconstruction result; generating a preset virtual surgical environment by a first generation unit based on the three-dimensional reconstruction result, simulating the motion trajectory, electromagnetic field distribution, and hydrodynamic parameters of the electromagnetic drive surgical robot in the preset virtual surgical environment, and obtaining first modal data according to the simulation result.

[0134] Optionally, in some embodiments, the second modal data includes at least one of the current of the electromagnetic coil, the magnetic field strength, the actual position of the surgical instrument, and the attitude of the surgical instrument.

[0135] Optionally, in some embodiments, the electromagnetic drive system further includes: determining the current of the electromagnetic coil by the hardware drive module according to the optimal control signal; generating a magnetic field strength by a plurality of electromagnetic coils of the electromagnetic coil module, and the plurality of electromagnetic coils generate a magnetic field strength based on the current of the electromagnetic coil; and collecting the actual position of the surgical instrument and the attitude of the surgical instrument of the electromagnetic drive surgical robot by the vision unit module.

[0136] Thus, through the joint training of the virtual simulation module and the electromagnetic drive system in the embodiments of the present application, high-precision, real-time, and intelligent autonomous control is achieved, which is applicable to various clinical tasks such as targeted drug release, in vivo sampling, and interventional therapy, showing broad adaptability and application potential.

[0137] It should be noted that the foregoing explanation of the embodiment of the autonomous control learning system of the electromagnetic drive surgical robot also applies to the autonomous control learning method of the electromagnetic drive surgical robot in this embodiment, and will not be elaborated here.

[0138] According to the autonomous control learning method of the electromagnetic drive surgical robot proposed in the embodiment of the present application, the virtual simulation module simulates the behavior of the electromagnetic drive surgical robot in a preset virtual surgical environment to obtain first modal data. The electromagnetic drive system receives the first modal data, and based on this data, through the actual operation of the electromagnetic coil module and the hardware drive module, obtains second modal data; the artificial intelligence calculation module generates an optimal control signal for the electromagnetic coil module based on the first modal data and, through a preset deep learning model and a preset diffusion model, to drive the electromagnetic drive surgical robot to autonomously complete tasks. Thereby, the problems of poor adaptability and insufficient flexibility of the electromagnetic drive surgical robot in the environment in the related art are solved, and the operation flexibility is improved.

[0139] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples.

[0140] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "N" is at least two, such as two, three, etc., unless otherwise specifically defined.

[0141] Any process or method description shown in the flowchart or described in other ways herein can be understood as representing a module, segment, or part of code including one or more executable instructions for implementing a customized logical function or process. The scope of the preferred embodiments of the present application includes additional implementations, where the functions can be executed in a substantially simultaneous manner or in a reverse order according to the functions involved, rather than in the order shown or discussed, which should be understood by those skilled in the art of the embodiments of the present application.

[0142] It should be understood that each part of the present application can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one of the following techniques known in the art or a combination thereof can be used: discrete logic circuits with logic gate circuits for implementing logical functions on data signals, application specific integrated circuits with suitable combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0143] Those of ordinary skill in the art can understand that all or part of the steps carried out in the method of the above embodiments can be completed by instructing relevant hardware through a program, and the said program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

Claims

1. An autonomous control learning system for an electromagnetic drive surgical robot, characterized in that Comprising: A virtual simulation module and an electromagnetic drive system. The electromagnetic drive system includes an electromagnetic coil module, a hardware drive module, and an artificial intelligence computing module. Among them, The virtual simulation module is used to simulate the behavior of the electromagnetic drive surgical robot in a preset virtual surgical environment to obtain first modal data; The electromagnetic drive system is used to receive the first modal data and, based on the first modal data, obtain second modal data through the actual operation of the electromagnetic coil module and the hardware drive module; The artificial intelligence computing module generates an optimal control signal for the electromagnetic coil module based on the first modal data and the second modal data through a preset deep learning model and a preset diffusion model, and drives the electromagnetic drive surgical robot to autonomously complete tasks based on the optimal control signal.

2. The autonomous control learning system of the electromagnetic drive surgical robot according to claim 1, wherein The virtual simulation module includes: A reconstruction unit for acquiring medical image data and, based on the medical image data, performing three-dimensional reconstruction on a human organ or tissue structure to obtain a three-dimensional reconstruction result; A first generation unit for generating the preset virtual surgical environment based on the three-dimensional reconstruction result, simulating the motion trajectory, electromagnetic field distribution, and hydrodynamic parameters of the electromagnetic drive surgical robot in the preset virtual surgical environment, and obtaining the first modal data according to the simulation results.

3. The autonomous control learning system of the electromagnetic drive surgical robot according to claim 1, characterized in that The second modal data includes at least one of the current of the electromagnetic coil, the magnetic field strength, the actual position of the surgical instrument, and the posture of the surgical instrument.

4. The autonomous control learning system of the electromagnetic drive surgical robot according to claim 3, characterized in that, The electromagnetic drive system further includes a visual unit module. Among them, The hardware drive module determines the current of the electromagnetic coil according to the optimal control signal; The electromagnetic coil module includes a plurality of electromagnetic coils, and the plurality of electromagnetic coils generate the magnetic field strength based on the current of the electromagnetic coil; The visual unit module is used to collect the actual position of the surgical instrument and the posture of the surgical instrument of the electromagnetic drive surgical robot.

5. The autonomous control learning system of the electromagnetic drive surgical robot according to claim 4, characterized in that, The electromagnetic drive system further includes: A wireless communication module for sending the second modal data to the virtual simulation module so that the virtual simulation module adjusts the simulation parameters of the preset virtual surgical environment according to the second modal data.

6. The autonomous control learning system of the electromagnetic drive surgical robot according to claim 1, wherein The artificial intelligence computing module includes: A receiving unit for receiving the first modal data and the second modal data; A processing unit for performing feature extraction and feature representation on the first modal data and the second modal data using the preset deep learning model, and planning the electromagnetic field based on the processing results through the preset diffusion model to obtain the optimal control signal.

7. An autonomous control learning method for an electromagnetic drive surgical robot, characterized in that, An autonomous control learning system for an electromagnetic drive surgical robot using any one of claims 1-6, wherein the method includes the following steps: Simulating the behavior of the electromagnetic drive surgical robot in a preset virtual surgical environment through the virtual simulation module to obtain first modal data; Receiving the first modal data through the electromagnetic drive system, and obtaining second modal data based on the first modal data through the actual operation of the electromagnetic coil module and the hardware drive module; Based on the first modal data and the second modal data, the artificial intelligence computing module generates an optimal control signal for the electromagnetic coil module through a preset deep learning model and a preset diffusion model, and drives the electromagnetic drive surgical robot to autonomously complete tasks based on the optimal control signal.

8. The autonomous control learning method of the electromagnetic drive surgical robot according to claim 7, characterized in that, The virtual simulation module simulates the behavior of the electromagnetic drive surgical robot in a preset virtual surgical environment to obtain first modal data, including: The reconstruction unit acquires medical image data, and based on the medical image data, performs three-dimensional reconstruction on a human organ or tissue structure to obtain a three-dimensional reconstruction result; The first generation unit generates the preset virtual surgical environment based on the three-dimensional reconstruction result, simulates the motion trajectory, electromagnetic field distribution, and hydrodynamic parameters of the electromagnetic drive surgical robot in the preset virtual surgical environment, and obtains the first modal data according to the simulation result.

9. The autonomous control learning method of the electromagnetic drive surgical robot according to claim 7, characterized in that, The second modal data includes at least one of the current of the electromagnetic coil, the magnetic field strength, the actual position of the surgical instrument, and the attitude of the surgical instrument.

10. The autonomous control learning method of the electromagnetic drive surgical robot according to claim 7, characterized in that, It further includes: The hardware drive module determines the current of the electromagnetic coil according to the optimal control signal; A plurality of electromagnetic coils of the electromagnetic coil module generate a magnetic field strength based on the current of the electromagnetic coil; The vision unit module collects the actual position of the surgical instrument of the electromagnetic drive surgical robot and the attitude of the surgical instrument.

Images