Permanent magnet array for magnetic control navigation system, control method and magnetic control navigation system

Abstract

The invention relates to the field of medical equipment, in particular to a permanent magnet array for a magnetic control navigation system, a control method and the magnetic control navigation system. The permanent magnet array comprises a plurality of permanent magnet type magnetic balls which are relatively fixed in position and adjustable in magnetic pole orientation; each magnetic ball is located on an arc-shaped curved surface in the vertical direction, and the permanent magnet array allows fine adjustment of the magnetic pole orientation of each magnetic ball through two motors which are independently configured. In practical application, the target magnetic field can be generated in the target area on the inner side of the arc-shaped curved surface only by jointly adjusting the magnetic pole orientation of each magnetic ball. In practical application, the preset path can be discretized into a plurality of control points, and the regulation and control instruction capable of following the expectation can be solved through quadratic programming. When the regulation and control strategy is solved, the optimization mode can be adjusted according to the actual working condition, so that the efficiency is improved while the safety is considered. The problems that existing equipment is insufficient in magnetic control precision and large in control difficulty are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of medical equipment, in particular to a permanent magnet array for a magnetic navigation system, a hybrid control method of the permanent magnet array and a magnetic navigation system. BACKGROUND

[0002] In recent years, with the gradual popularization of vascular intervention, surgical robot technology has broken through, and has promoted the precision and safety of vascular intervention. Domestic and foreign institutions have made significant progress in the field of neural interventional magnetic navigation systems, which are mainly used for navigation and execution of specific medical functions in the human body. Compared with traditional manual control, the magnetic control method has many advantages such as remote operation, high precision, strong control, non-invasive treatment, strong biocompatibility, and no harm to the human body. However, due to the limitations of the neural interventional magnetic navigation system such as large space, insufficient magnetic field strength, non-uniform magnetic field, and difficult magnetic field regulation, the application potential of the magnetic navigation system in human vascular intervention navigation has not been fully developed.

[0003] At present, the neural interventional magnetic navigation system on the market mainly generates excitation magnetic field by electromagnet or permanent magnet. The electromagnet generates magnetic field by energizing the coil. These coils are usually solenoid-shaped and are wound around a magnetic material with low resistance material. When the coil is energized, the current generates a magnetic field in it. Because the current flowing through the wire with resistance will generate heat, so in the design of electromagnet, cooling device needs to be included to prevent overheating. At the same time, in order to prevent current short circuit, insulating material must be used between the coils. The electromagnet magnetic navigation system can adjust the strength of the magnetic field by changing the size of the current and the number of turns of the external coil. By deploying multiple electromagnets in different directions, the adjustable function of the magnetic field direction can be realized. However, it also has many problems, such as large volume and weight, especially when a strong magnetic field is needed, the large volume usually brings problems such as long operating room modification period and poor physician intervention.

[0004] In addition to electrically generated magnetic field, permanent magnet is another commonly used material. By fixing permanent magnet at the end of the mechanical arm actuator, and using the attitude control of the mechanical arm, the magnetic field of the target position can be adjusted. However, the permanent magnet type magnetic navigation system also has problems, such as the neural interventional magnetic navigation system based on mechanical arm driving a single permanent magnet. Due to the non-uniform distribution of the external magnetic field of the permanent magnet, and the rapid decline of the magnetic field strength in space with the increase of the distance, the system has the problems of low magnetic field strength and poor control precision. Because the magnetic field of the permanent magnet cannot be permanently closed, the neural interventional magnetic navigation system based on permanent magnet is prone to non-ideal intermediate magnetic field during different magnetic field switching, which disturbs the magnetic controlled object and increases the uncontrollability of the operation. SUMMARY

[0005] In order to solve the problems of insufficient precision, great control difficulty and irrational intermediate magnetic field existing in the prior art permanent magnet type magnetic control navigation system, the application provides a permanent magnet array for a magnetic control navigation system and a corresponding hybrid control method of the permanent magnet array and a magnetic control navigation system.

[0006] The technical scheme provided by the application is as follows: A permanent magnet array for a magnetic control navigation system comprises a base, a magnetic sphere array and a magnetic pole adjusting mechanism.

[0007] The base comprises an arc-shaped curved surface in the vertical direction. The magnetic sphere array is installed on the arc-shaped curved surface of the base by a support and is arranged in a 3*3 array in the vertical space of the arc. Each magnetic sphere in the magnetic sphere array is rotatably connected to the support by a rotating shaft perpendicular to the magnetic domain direction, and the rotating shaft is arranged in the horizontal direction. The support is rotatably connected to the base. The three magnetic spheres in each layer of the magnetic sphere array are located at the and polar angles of the same horizontal direction inscribed circle, wherein , the diameter of the inscribed circle is k d times the diameter of the magnetic sphere, the height between the layers of the magnetic spheres is k h times the diameter of the magnetic sphere, .

[0008] The magnetic pole adjusting mechanism comprises two rotatable joints of each magnetic sphere, and the two joints are respectively composed of a first motor and a second motor installed between the magnetic sphere and the support. The first motor is used to drive the combination of the support and the magnetic sphere to rotate in the direction perpendicular to the arc-shaped curved surface, and the second motor is used to drive the magnetic sphere to spin along the rotating shaft. When the rotation angles of the motors in the two joints change, the magnetic pole orientation of the corresponding magnetic sphere will change.

[0009] In the permanent magnet array of the application, the relative positions of each magnetic sphere are fixed, the first motor and the second motor are jointly controlled to adjust the magnetic pole orientation of each magnetic sphere, and then the magnetic field direction and magnetic field strength of the target area on the inner side of the arc-shaped curved surface are changed.

[0010] The application also comprises a hybrid control method of a permanent magnet array, which is used to cooperatively control the postures of each magnetic sphere in two mirror image arranged permanent magnet arrays in a magnetic control navigation system, so as to make the magnetic field states of each control point on the preset path in the target area follow the expectation, so as to realize the movement of the magnetic control target along the preset path. The control method provided by the application comprises the following steps: One, the magnetic field state of 8 degrees of freedom is constructed by using the magnetic field strength components of each point in the target area in three orthogonal directions and the magnetic induction strength gradient in five different directions as the state dimensions U ; the rotation angles of the two joints of each magnetic ball are taken as its attitude, and the 36-dimensional spatial attitude of the magnetic control navigation system is constructed q ; the magnetic control mapping representing the magnetic field state in the target area under different array attitudes is constructed in advance: U = J ( q )· q , wherein J ( q ) is the magnetic Jacobian matrix used to realize the conversion.

[0011] Two, the preset path is adaptively discretized into N consecutive control points combined with the trajectory curvature and the singular risk S s k : ; and the magnetic field state expectation of each control point is preset . Wherein, N represents the number of control points on the preset path S .

[0012] At each time t , the optimal adjustment action satisfying the constraint is generated according to the current magnetic field state s k of the control point U k and the preset magnetic field state expectation of the next control point a t , and then the magnetic field following on the preset path is realized through the optimal adjustment action sequence A: A={ a 1、 a 2… a t}; wherein, the generation strategy of the optimal adjustment action a t at each time is as follows: 2.1, the risk index under the current state is calculated based on the Jacobian matrix J ( q ) and the system constraint , and the optimization mode is specified: When is higher than the preset risk threshold , the parallel mode of synchronous optimization of all degrees of freedom is adopted; otherwise, the sequential mode of sequential optimization of each degree of freedom is adopted.

[0013] ​​In the sequential mode, each degree of freedom is taken as a node, and the edge weight is set according to the crosstalk of each degree of freedom to other degrees of freedom, and a sequential cost graph is constructed. In the sequential cost graph, a loop-free path π is searched out which can traverse all nodes and makes the total path cost minimum, and the order of nodes in π is taken as the optimization order of each degree of freedom.

[0014] 2.2, minimizing the action penalty P For the optimization target, the joint speed of each magnetic ball which can follow the magnetic field state expectation and meet various physical constraints is solved by quadratic programming according to the specified optimization mode .

[0015] 2.3, according to the preset action time The Euler integral is performed on the joint speed to calculate the spatial pose q of the magnetic ball array after executing the regulation action t+1 : ; in combination with and q t+1 , the optimal regulation action a t is generated, and the regulation instruction is issued to the motor of each magnetic ball.

[0016] As a further improvement of the application, a neural network-based magnetic field simulation agent model is constructed and trained in advance, which is used to simulate the magnetic control mapping of the magnetic field state in the target region under different array poses.

[0017] An optimization mode agent model based on graph neural network is constructed, which is used to take each degree of freedom as a node, set the edge weight according to the crosstalk of each degree of freedom to other degrees of freedom, and construct a sequential cost graph; in the sequential cost graph, a loop-free path π is searched out which can traverse all nodes and makes the total path cost minimum, and the order of nodes in π is taken as the optimization order of each degree of freedom in the sequential optimization mode.

[0018] An agent model based on reinforcement learning is constructed, which is used to minimize the action penalty P as the optimization target, and output the joint speed of each magnetic ball which can follow the magnetic field state expectation and meet various physical constraints according to the specified optimization mode ; and according to the preset action time , the Euler integral is performed on the joint speed to calculate the spatial pose q of the magnetic ball array after executing the regulation action t+1 : ; in combination with and q t+1 , the optimal regulation action a t is generated; The control method of the permanent magnet array as described above is realized by using a federation model comprising a magnetic field simulation agent model, an optimization mode agent model and a planning decision agent model.

[0019] The application also comprises a magnetic navigation system comprising two mirror arranged magnetic control machines and a rack capable of driving the lifting movement of the permanent magnet array; the magnetic control machine comprises the permanent magnet array for the magnetic navigation system as described above.

[0020] The magnetic navigation system also comprises a controller; the controller uses the hybrid control method of the permanent magnet array as described above to cooperatively control the two mirror arranged permanent magnet arrays in the magnetic navigation system, and by issuing instructions to the two motors of each magnetic ball, the magnetic field intensity of each control point in the target area follows the expectation to realize the movement of the magnetic control target along the preset path. S The magnetic field intensity of each control point s k follows the expectation to realize the movement of the magnetic control target along the preset path.

[0021] The application has the following beneficial effects: The application provides a novel structure and control logic of a permanent magnet array of a magnetic navigation system, wherein the permanent magnet array in the scheme is arranged in a vertical plane by a plurality of magnetic balls. The positions of the magnetic balls in the permanent magnet array are relatively fixed, and the magnetic pole orientation of the magnetic balls can be flexibly adjusted by two independent joint motors, thereby realizing fine control of the magnetic field state in the target space. The control difficulty of the new permanent magnet array is relatively simple, the collision risk of the traditional mechanical arm control scheme can be overcome, the magnetic field control range in the target area is larger, the continuity is better, more flexible magnetic field control can be realized, and the device energy consumption can be effectively reduced.

[0022] For the new permanent magnet array provided by the application, the application further designs a corresponding control strategy. The control strategy of the application can flexibly adjust the solution of the effective solution under different working conditions, thereby greatly improving the efficiency of generating optimized control actions on the basis of ensuring safety and ensuring the real-time performance of the magnetic navigation system. The application also introduces a machine learning algorithm in the control scheme, thereby reducing the requirement of the magnetic navigation system on the device computing power through the hybrid control logic; and improving the practical value of the scheme. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 FIG. 1 is a structural schematic diagram of the permanent magnet array for the magnetic navigation system provided in embodiment 1 of the application.

[0024] Figure 2 FIG. 3 is a spatial layout diagram of each magnetic ball in each layer of the permanent magnet array in embodiment 1 of the application.

[0025] Figure 3Fig. 1 is a structural schematic diagram of the permanent magnet array in Example 1 of the present application from one perspective.

[0026] Figure 4 Fig. 2 is a structural schematic diagram of the permanent magnet array in Example 1 of the present application from another perspective.

[0027] Figure 5 Fig. 3 is a structural schematic diagram of the magnetic navigation system in Example 2 of the present application.

[0028] Figure 6 Fig. 4 is a flow chart of the control method adopted by the controller of the magnetic navigation system in Example 2 of the present application.

[0029] Figure 7 Fig. 5 is a principle diagram of generating control points and magnetic field state expectations according to a preset trajectory in Example 2 of the present application.

[0030] Figure 8 Fig. 6 is a cost sequence diagram for generating a loop-free path constructed in Example 2 of the present application.

[0031] The reference signs in the figures are: 1, base; 2, magnetic ball array; 3, magnetic pole adjusting mechanism; 4, permanent magnet array; 5, magnetic control machine; 6, machine frame; 21, magnetic ball; 31, base; 32, support; 33, first motor; 34, second motor; 35, rotating shaft; 36, gear set. DETAILED DESCRIPTION

[0032] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0033] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

[0034] Example 1 In view of the problems of low magnetic field strength and poor control accuracy of the existing permanent magnet-based magnetic control navigation system controlled by a mechanical arm, the embodiment provides a permanent magnet array 4 for a magnetic control navigation system, which is a magnetic head of each magnetic control machine 5 in the magnetic control navigation system and can determine the control accuracy of the magnetic field output by the magnetic control navigation system. Unlike the traditional scheme of adjusting the position and attitude of the permanent magnet to change the magnetic field of the target area by a mechanical arm, the permanent magnet array 4 provided by the embodiment comprises a plurality of permanent magnet type magnetic balls 21 with relatively fixed positions but adjustable magnetic pole orientations; the permanent magnet array 4 allows fine adjustment of the magnetic pole orientations of each magnetic ball 21 by two independently configured motors. In actual application, the embodiment can only adjust the magnetic pole orientations of each arrayed magnetic ball 21 to generate a dynamically changing magnetic field required for a magnetic control traction task in a target area; this can greatly reduce the difficulty of control and improve the control accuracy of the magnetic field.

[0035] Specifically, as shown in the drawings, Figure 1 The permanent magnet array 4 provided by the embodiment is composed of a base 1, a magnetic ball array 2, and a magnetic pole adjusting mechanism 3. The base 1 includes an arc-shaped curved surface in the vertical direction. The magnetic ball array 2 comprises nine permanent magnet type magnetic balls 21 mounted on the arc-shaped curved surface of the base 1 through a support 32 and arranged in a 3x3 array in the vertical space of the arc-shaped surface. Each magnetic ball 21 in the magnetic ball array 2 is rotatably connected to the support 32 through a rotation shaft 35 perpendicular to the magnetic domain direction, and the rotation shaft 35 is arranged in the horizontal direction; the support 32 is rotatably connected to the base 1.

[0036] In particular, considering that the scheme of the embodiment only supports fine adjustment of the magnetic pole orientations of each magnetic ball 21, the relative positions of each magnetic ball 21 remain fixed. Although this design can reduce the complexity of magnetic field control in the target area, it may also affect the flexibility of the control means to some extent and reduce the feasible region of the magnetic field state distribution to some extent. In view of this problem, the relative positions of each permanent magnet array 4 are optimized in the embodiment, so that each permanent magnet is in the best spatial position, which can maximize the adjustable range of the magnetic field of the permanent magnet array 4 and enable the magnetic field change at any point in the target area to be continuous without sudden changes.

[0037] In the embodiment, on the one hand, the overall form of the magnetic ball array 2 is optimized and controlled so that each magnetic ball 21 in the magnetic ball array 2 is arranged in an arc-shaped curved surface. Based on the superposition principle of the magnetic field, this structure design can form a stronger target magnetic field in the space on the inward side, thereby laying a foundation for fine magnetic field control and widening the adjustable range of the magnetic field strength of each point in the target space. On the other hand, the relative positions of each magnetic ball 21 are optimized. Figure 2As shown in the actual application of the embodiment, the three magnetic spheres 21 of each layer of the magnetic sphere array 2 are located at the and polar angles of the same horizontal direction inscribed circle , . In addition, the diameter of the inscribed circle is k d times the diameter of the magnetic sphere 21, and the height between the layers of the magnetic sphere 21 is k h times the diameter of the magnetic sphere 21. In the optimized scheme of the embodiment, , , ; Under the optimal parameters provided by the embodiment, the magnetic field regulation effect of the target region is best, and there is no singular point.

[0038] The direction of the external magnetic field generated by the permanent magnet is always from the N pole to the S pole; therefore, by adjusting the direction of the magnetic pole of the permanent magnet, the magnetic field strength at any point around the permanent magnet can be changed. In the embodiment, the reason for using the magnetic sphere 21 type permanent magnet is mainly to achieve more fine adjustment of the magnetic pole direction of the permanent magnet and expand the adjustable range of the magnetic pole direction angle. The magnetic domain direction (the direction of the magnetic pole) of the magnetic sphere 21 is located on one of the diameters, and the magnetic pole adjusting mechanism 3 of the embodiment includes two rotatable joints of each magnetic sphere 21, and the two joints are respectively composed of a first motor 33 and a second motor 34 installed between the magnetic sphere 21 and a bracket 32 thereof. The first motor 33 is used to drive the combination of the bracket 32 and the magnetic sphere 21 to rotate in a direction perpendicular to the arc surface as the axis, and the second motor 34 is used to drive the magnetic sphere 21 to spin along the rotation axis 35. When the rotation angles of the motors in the two joints change, the magnetic pole direction of the corresponding magnetic sphere 21 will change.

[0039] Compared with traditional large magnets, the effective working area of the magnetic sphere array 2 of the embodiment is larger. This characteristic is particularly important in various types of interventional fine surgery scenes. When the target point is close to a certain magnetic sphere 21, other magnetic spheres 21 are still in the far field region, and the magnetic field distribution can still be accurately predicted by the magnetic dipole model, thereby reducing the prediction error. Through array combination and spatial optimization layout, when several magnetic spheres 21 are in the near field range of the target point, the remaining magnetic spheres 21 in the far field state can still provide high-precision magnetic field components, and the global prediction deviation is reduced through error mutual offset. This distributed magnetic source design provides a basis for accurate magnetic field modeling and flexible regulation, thereby significantly improving the control accuracy and dynamic performance of the navigation system.

[0040] In the actual application of the embodiment, the structure of the magnetic pole adjusting mechanism 3 is as Figure 3 andFigure 4 As shown, it includes a base 31, a support 32, a first motor 33, a rotating shaft 35, a gear set 36 and a second motor 34. Specifically, the support 32 is rotatably connected with the base 31, and the first motor 33 is fixedly connected with the base 31, with its output shaft fixedly connected with the support 32. The magnetic ball 21 is rotatably connected with the outward side of the support 32 through a penetrating or two-section rotating shaft 35, while the second motor 34 is fixedly installed on the inward side of the support 32, with its output shaft drivingly connected with the rotating shaft 35 of the magnetic ball 21 through the gear set 36. As shown, Figure 3 When the second motor 34 operates, its output shaft drives the rotating shaft 35 to rotate through the gear set 36, thereby causing the magnetic ball 21 to spin. When the first motor 33 operates, its output shaft drives the combination of the magnetic ball 21, the support 32, the rotating shaft 35, the gear set 36 and the second motor 34 to rotate relative to the base 31.

[0041] As can be seen, in the scheme of the present embodiment, the rotating directions of the movable joint formed by the two motors and the magnetic domain direction of the magnetic ball 21 are located in three different orthogonal directions. Therefore, by finely controlling the rotation angles of the two motors, the magnetic pole of the magnetic ball 21 can be oriented in any direction within the spherical space. And ultimately, on the basis of the fixed relative position of each magnetic ball 21, the steering angles of the first motor 33 and the second motor 34 are jointly controlled to adjust the magnetic pole orientation of each magnetic ball 21, thereby changing the magnetic field direction and magnetic field strength of the target region on the inward side of the arc-shaped curved surface.

[0042] In addition, it should be additionally noted that, in actual applications, in order to avoid interference with the magnetic field generated by the magnetic ball 21, the components in the present embodiment, such as the protective shell, the base 1, the base 31, the support 32, the rotating shaft 35 and the gear set 36, should be made of non-ferromagnetic materials. For example, the base 1 and the protective shell can be made of resin materials, while the rotating shaft 35 and the gear set 36 can be made of ceramic materials.

[0043] Embodiment 2 Based on the permanent magnet array 4 with a brand-new structure and a new type of control mode provided in Embodiment 1, the present embodiment further provides a corresponding magnetic control navigation system, as shown in Figure 5As shown, the magnetic navigation system includes two mirror arranged magnetic control machines 5, and a rack 6 fixedly connected with each magnetic control machine 5 and capable of driving the permanent magnet array 4 in the magnetic control machine 5 to move up and down as a whole. In actual application, the rack 6 can adopt various lifting tables or multi-degree-of-freedom robots. In particular, the magnetic control machine 5 includes the permanent magnet array 4 for the magnetic navigation system as in Embodiment 1 and a protective shell wrapping the permanent magnet array 4. In the magnetic control machine 5, each magnetic ball 21 in the permanent magnet array 4 faces a magnetic field generating unit that can be independently controlled, and each magnetic field generating unit arranged in a 3x3 array on both sides can synthesize the required target magnetic field at the target region in the middle.

[0044] In order to realize joint control of the spatial poses of each magnetic ball 21 in the permanent magnet array 4 of the two magnetic control machines 5, the magnetic navigation system of the embodiment further includes a controller. The controller is used to cooperatively control the two mirror arranged permanent magnet arrays 4 in the magnetic navigation system, change the pole direction of each magnetic ball 21 by issuing instructions to the two motors of each magnetic ball 21, and further make the magnetic field strength of a certain preset path in the target region dynamically change as expected to realize the movement of the magnetic control target along the preset path. S The above control points s k The magnetic field strength of the above control points

[0045] Specifically, as Figure 6 shown, the control strategy adopted by the controller in the magnetic navigation system of the embodiment is as follows: S1: Construct the mapping relationship between the spatial pose q of the magnetic navigation system and the magnetic field state in the target region.

[0046] In the magnetic navigation system of the embodiment, the magnetic field state in the target region between the two magnetic control machines 5 is generated by each magnetic ball 21. For any point in the target region, the magnetic field strength and direction of the point are determined by the vector superposition of the magnetic fields of the 18 magnetic balls 21 in the two permanent magnet arrays 4 at this point. Therefore, a corresponding mapping relationship can be established between the poses of each magnetic ball 21 in the magnetic navigation system and the finally synthesized magnetic field of the target region, so as to lay a foundation for the subsequent accurate realization of the magnetic field regulation target.

[0047] In the embodiment, the magnetic field B ( r ) generated by a single magnetic ball 21 at various positions in space can be calculated by the following magnetic dipole model: ; In the above formula, m represents the magnetic moment of the permanent magnet magnetic ball 21, r represents the position vector of the target point in space relative to the permanent magnet; and It is the vacuum permeability.

[0048] Correspondingly, under the influence of the two permanent magnet arrays 4 in the magnetically controlled navigation system, the total magnetic field at any point in space... B ( r The result can be calculated using the following magnetic field superposition model: ; In the above formula, m i For the first i The magnetic moment direction of each permanent magnet (controlled by the rotational degree of freedom driven by two motors); r i For the target point relative to the first i The position vectors of the permanent magnets. In practical applications, the above magnetic field calculation process can be quickly completed using Python in conjunction with the magicylib library.

[0049] Considering the complex coupling relationships between the magnetic fields generated by different permanent magnets, in order to achieve precise control of the magnetic field in the target region, this embodiment uses the magnetic field intensity components in three orthogonal directions and the magnetic induction intensity gradients in five different directions at each point in the target region as state dimensions to characterize the magnetic field state of the target domain. U This allows for the construction of corresponding 8-degree-of-freedom vectors representing the magnetic field state. In practical control, each degree of freedom of each magnetic field state can be simultaneously optimized to find a strategy that can achieve the state in one go; alternatively, each degree of freedom can be approximated sequentially to gradually find a step-by-step strategy that can achieve the magnetic field state.

[0050] Specifically, the data format of the magnetic field state U vector in this embodiment is as follows:

[0051] in, B x , B y and B z These represent the magnetic field strength components along the x-axis, y-axis, and z-axis, respectively. G xx , G xy , G xz , G yy and G yz These represent the magnetic flux density gradients in five different directions; the calculation formulas are as follows: , , , and .

[0052] Correspondingly, the magnetic field state U of the target region in the embodiment is determined by the magnetic pole orientation of each magnetic ball 21 in the magnetic control navigation system, and the magnetic pole orientation of each magnetic ball 21 can be controlled through two active joints. Therefore, in order to facilitate calculation, the embodiment can directly represent the posture of each magnetic ball 21 through the rotation angle of the two joint motors of the magnetic ball 21. And on this basis, a 36-dimensional spatial posture is constructed, which can represent the overall posture of the two permanent magnet arrays 4 of the magnetic control navigation system. The posture vector of the spatial posture of the magnetic control navigation system is denoted as q , .

[0053] In combination with the foregoing description, based on the principle of magnetic field superposition, the embodiment can determine the magnetic field state U generated by the magnetic control navigation system in the target region at any spatial posture q through an analytical and simulation method. Therefore, the embodiment can finally construct a magnetic control mapping representing the magnetic field state in the target region under different array postures, which is denoted as: U = J ( q )· q , wherein, J ( q ) is a magnetic Jacobian matrix used for conversion.

[0054] S2: decompose the continuous control task of the motion trajectory of the magnetic control target into a task of regulating the magnetic field state of each discrete control point on the preset path, and optimize each regulation action in the process. Specifically, the content of the above process includes: S21: combine the trajectory curvature and singularity risk to adaptively discretize the preset path S into a plurality of continuous control points s k : , and preset the magnetic field state expectation of each control point. Wherein, N represents the number of control points on the preset path S . As shown in Figure 7 , in actual application, when the curvature of a certain interval on the preset path is large, the number of control points for the interval should be increased. When the singularity risk of some regions is high, it means that there is a risk of exceeding the rotation angle of the motor, and the number of control points in these intervals should also be increased. In addition, the control difficulty of other regions is relatively low, and the control points can be set at a standard distance.

[0055] By discretizing the motion trajectory control task, this embodiment can gradually adjust the magnetic field state at each point on the preset path to match the desired value within a series of consecutive action times, thereby gradually achieving the goal of guiding the magnetically controlled target to move along the preset path. Specifically, at each moment... t According to control points s k Current magnetic field state U k And the expected magnetic field state of the next control point. Generate the optimal adjustment action that satisfies the constraints. a t Furthermore, by optimizing the action sequence A: A = { a 1. a 2… a t It enables magnetic field following on a preset path.

[0056] S22: In practical applications, at each moment t Optimal adjustment action a t The generation strategy is as follows: Based on Jacobian matrix J ( q ) and system constraints to calculate risk indicators under the current state And specify the optimization mode: Risk indicators This is an indicator used to assess the risk of switching the current magnetic field state to a preset desired magnetic field state under different conditions; it is related to multiple factors such as the condition number of the control process, joint movement speed, and magnetic field gradient. As mentioned earlier, there is a correlation between different degrees of freedom in the magnetic field state. Adjusting one often has a correlated impact on other degrees of freedom. Therefore, when optimizing control actions under different conditions, different strategies should be adopted for each degree of freedom of the target magnetic field state to achieve the desired result. For example, when the risk index is at a low level, in order to increase the rate, while ensuring that no over-limit events or violations of physical constraints occur, it is advisable to optimize each degree of freedom of the magnetic field state in one go. However, when the risk index is at a high level, in order to ensure the feasibility of the control action and avoid violating physical constraints, the solution space that can satisfy the requirements of each degree of freedom of the magnetic field state can be searched one by one, and finally the optimal control action that can follow the desired magnetic field state can be found.

[0057] In practical applications of this embodiment, risk indicators The calculation formula is as follows: ; In the above formula, denotes a condition number normalization index related to J . denotes joint velocity; denotes magnetic field change rate; denotes current condition number; denotes a reference condition number for normalizing to a reference scale; H B denotes derivative of magnetic field to joint; denotes damping pseudo-inverse; denotes upper limit of joint velocity; denotes first derivative of process magnetic field state expectation U . ref B denotes magnetic field strength; t denotes time; denotes upper limit of magnetic field slope; denotes q first derivative, i.e. joint velocity.

[0058] The adaptive selection of optimization strategy according to risk index is as follows: When is higher than a preset risk threshold , a parallel mode of synchronous optimization of all degrees of freedom is adopted; otherwise, a sequential mode of sequential optimization of each degree of freedom is adopted.

[0059] In the sequential mode, considering the complex coupling relationship among different degrees of freedom in the magnetic field state, the order of each degree of freedom in the solving process also needs to be further optimized. In this embodiment, the strategy for optimizing the order of each degree of freedom in the sequential mode is as follows: S23: Taking each degree of freedom as a node, setting edge weight by combining the crosstalk of each degree of freedom adjustment to other degrees of freedom and constructing a sequential cost graph. In the sequential cost graph, search for a loop-free path π that can traverse all nodes and make the total path cost minimum, and take the order of nodes in π as the optimization order of each degree of freedom.

[0060] In detail, in this embodiment, as shown in Figure 8 , the search method of path π is as follows: (1) Taking each degree of freedom as a node and the influence of each degree of freedom on other degrees of freedom during adjustment as an edge, a directed acyclic graph G is constructed.

[0061] (2) When only the arbitrary i th degree of freedom is excited from the current value to the expected value, the other jMaximum state change in each degree of freedom The normalized value is used as a crosstalk index Construct a cross-activation matrix C , : .

[0062] (3) Combining crosstalk index Setting diagram G any directed edge in i → j edge weight g ij The following is a sequence cost graph: ; In the above formula, SatRatio ( q ) indicates and q The reciprocal measure of the relevant minimum attitude limit margin; RateRatio This represents the ratio of the rate of change of the magnetic field to the upper limit of the rate of change of the gradient; w 1. w 2. w 3. w 4 represents the weights of the four indicators.

[0063] (4) Based on the edge weights of the sequential cost graph g ij We use dynamic programming or integer programming to search for an acyclic path π that can traverse all nodes and minimizes the total cost of the path; where an acyclic path is defined as an optimized path that does not contain any cycles and any node is visited only once in the entire path.

[0064] S24: After determining the optimization mode, the control action generation task of the magnetic navigation system in this embodiment is essentially a multi-objective optimization task. To address this issue, this embodiment minimizes action penalties. P To optimize the objective, following the specified optimization model, quadratic programming is used to solve for the joint velocities of each magnetic ball 21 that can follow the desired magnetic field state and satisfy all physical constraints. Then, adjust the joint speed according to the preset motion time. Perform Euler integration to calculate the spatial attitude q of magnetic sphere array 2 after the control action is performed. t+1 : ; combination and q t+1 Generate optimal adjustment action a t It also sends adjustment commands to the motors of each magnetic ball 21.

[0065] Specifically, in the secondary planning of this embodiment, action penaltyP The expression is as follows: ; The five penalty terms contained in the above formula are in turn the tracking penalty, the action amplitude penalty, the spatial overrun penalty, the magnetic field climb penalty and the gradient climb penalty. Among them, the first penalty term is used to ensure that the 8 degrees of freedom change according to the reference; the second penalty term is used to suppress the action amplitude; the third penalty term is used to realize the avoidance of the limit position and decoupling in the null space; the fourth and fifth penalty terms are used to limit the climbing rate of the magnetic field and the gradient; W U , W q , W N , W B , W G are the weight matrices of each penalty term respectively; respectively represent the influence weights of the last four penalty terms; I represents the unit matrix; represents the derivative of the gradient to the joint.

[0066] The physical constraints that each control action performed by the magnetic control navigation system in the regulation process of the embodiment needs to meet mainly include multiple levels such as joint rotation angle, joint speed, joint acceleration, change rate of joint acceleration, change rate of magnetic field strength and gradient, etc. Roughly include: , , , , ; and q need to meet the limit position constraint of the joint position and the spatial constraint of avoiding collision.

[0067] Among them, respectively represent the second derivative and the third derivative of q , that is, the joint acceleration and the change rate of the joint acceleration; q velmax , q accmax and q jerkmax respectively represent the upper limit of the joint speed, the joint acceleration and jerk ; represents the upper limit of the gradient slope.

[0068] In addition, the foregoing risk index is merely a rough index for adaptive switching of optimization modes for each degree of freedom in the magnetic field state. Even if the risk index is less than the preset risk threshold, there can still be a case that is not applicable to the parallel mode, so the present embodiment also sets a corresponding daemon and fallback mechanism for the foregoing scheme. Specifically, when the optimization result that satisfies the constraints cannot be solved in any quadratic programming based on the parallel mode, the present embodiment scheme can return to the sequential mode to re-solve. Further, when the optimization result that satisfies the constraints cannot be solved in any quadratic programming based on the sequential mode, the control points in the corresponding interval of the motion trajectory s k is encrypted or the preset action time is extended, and then re-solved.

[0069] wherein the time expansion factor adopted by the action time extension is automatically contracted according to the tightest constraint, and satisfies the following formula: ; In the above formula, is a conservative coefficient, ; is the 8-degree-of-freedom difference from the current state to the target state, and is determined by the path derivative and the current path direction determines its time scale.

[0070] Embodiment 3 Embodiment 2 provides a control logic for generating an optimized control strategy based on an analytical method for the controller in the magnetic control navigation system, which may be limited by the computing power of the controller in actual application and may not be real-time enough. To solve this problem, the present embodiment further provides a hybrid control method by introducing a machine learning algorithm. Specifically, in the present embodiment, a neural network-based magnetic field simulation agent model is constructed and trained in advance, which is used to simulate the magnetic control mapping of the magnetic field state in the target region under different array postures.

[0071] An optimization mode agent model based on a graph neural network is constructed in advance, which is used to set edge weights by combining the crosstalk of each degree of freedom adjustment on other degrees of freedom and construct a sequential cost graph with each degree of freedom as a node. In the sequential cost graph, a loop-free path π that can traverse all nodes and makes the total path cost minimum is searched out, and the order of nodes in π is taken as the optimization order of each degree of freedom in the sequential optimization mode.

[0072] An agent model for planning and decision-making based on reinforcement learning is constructed in advance, which is used to minimize the action penalty PFor optimization target, according to the specified optimization mode, output the joint speed of each magnetic ball 21 that can follow the magnetic field state expectation and meet various physical constraints ; and according to the preset action time , Euler integral is performed on the joint speed , and the spatial pose q of the magnetic ball array 2 after executing the regulation action is calculated t+1 : ; and then the optimal regulation action is generated in combination with t+1 q a t ; The federated model including the magnetic field simulation agent model, the optimization mode agent model and the planning decision agent model is adopted to realize the control method of the permanent magnet array 4 based on the safety area as described above. In actual application, after the three trained agent models are deployed in the controller of the magnetic control navigation system, the three agent models can be used to assist in completing the large amount of complex matrix operation tasks contained in the control method in Embodiment 2, thereby greatly reducing the operation load of the controller, improving the data processing efficiency of the controller, and ensuring the real-time performance of the magnetic control navigation system.

[0073] The above-described embodiments only express one of the embodiments of the present application, which is described in detail and in detail, but it cannot be understood as a limitation on the scope of the application. It should be noted that for ordinary skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are within the scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A permanent magnet array for a magnetically controlled navigation system, characterized in that, It includes: The base includes an arcuate surface along the vertical direction; A magnetic sphere array, consisting of nine permanent magnet-shaped magnetic spheres arranged in a 3×3 array, is mounted on the arc-shaped surface of the base via a support. Each magnetic sphere is rotatably connected to the support via a pivot perpendicular to the magnetic domain direction, the pivot being arranged horizontally. The support and the base are rotatably connected. In each layer of the magnetic sphere array, three magnetic spheres are located within the same horizontally oriented inscribed circle. and At the polar angle; among them, , The diameter of the inscribed circle is equal to the diameter of the magnetic sphere. k d times, The interlayer height of the magnetic spheres is equal to the diameter of the magnetic spheres. k h times, ; A magnetic pole adjustment mechanism includes a joint consisting of a first motor and a second motor installed between each magnetic ball and its support; the first motor drives the assembly of the support and the magnetic ball to rotate about an axis perpendicular to the arcuate surface; the second motor drives the magnetic ball to spin along the axis of rotation. The relative position of each magnetic ball in the permanent magnet array is fixed. The first and second motors are jointly controlled to adjust the orientation of the magnetic poles of each magnetic ball, thereby changing the magnetic field direction and magnetic field strength of the target area on the inner side of the arc-shaped surface.

2. The permanent magnet array for a magnetically controlled navigation system as described in claim 1, characterized in that, In the spatial arrangement of the magnetic sphere array, the optimal parameters are: , , , .

3. A hybrid control method for a permanent magnet array, characterized in that: It is used to coordinate the attitude of each magnetic ball in two mirror-arranged permanent magnet arrays in a magnetic navigation system, so that the magnetic field state of each control point on the preset path in the target area follows the expectation, so as to realize the manipulation of the magnetically controlled target to move along the preset path. The control method includes: An 8-degree-of-freedom magnetic field state is constructed, with the magnetic field intensity components at each point within the target region in three orthogonal directions and the magnetic induction gradients in five different directions serving as the state dimensions. U The rotation angles of the two joints of each magnetic sphere are used as its attitude, thereby constructing the 36-dimensional spatial attitude of the magnetically controlled navigation system. q ;Pre-construct a magnetocontrol mapping that characterizes the magnetic field state within the target region under different array attitudes: U = J ( q )· q ,in, J ( q ) is the magnetic Jacobian matrix used to implement the transformation; Combining trajectory curvature and singularity risk to pre-set path S Adaptive discretization into N continuous control points s k : And preset the desired magnetic field state for each control point; ; at every moment t According to control points s k Current magnetic field state U k And the expected magnetic field state of the next control point. Generate the optimal adjustment action that satisfies the constraints. a t Furthermore, by optimizing the action sequence A: A = { a 1. a 2… a t } To achieve magnetic field following on a preset path; where, a t The generation strategy is as follows: Based on Jacobian matrix J ( q ) and system constraints to calculate risk indicators under the current state And specify the optimization mode: when Risk threshold higher than preset If all degrees of freedom are optimized simultaneously, a parallel mode is adopted; otherwise, a sequential mode is adopted to optimize each degree of freedom in turn. In the sequential mode, each degree of freedom is used as a node, and edge weights are set in combination with the crosstalk of each degree of freedom to other degrees of freedom when adjusting each degree of freedom, and a sequential cost graph is constructed. In the sequential cost graph, the acyclic path π that can traverse all nodes and minimizes the total cost of the path is searched, and the order of the nodes in π is used as the optimization order of each degree of freedom. To minimize action penalties P To optimize the objective, following a specified optimization model, quadratic programming is used to solve for the joint velocities of each magnetic sphere that can follow the desired magnetic field state and satisfy all physical constraints. According to the preset action time For joint velocity Perform Euler integration to calculate the spatial attitude q of the magnetic sphere array after the control action is performed. t+1 : ; combination and q t+1 Generate optimal adjustment action a t It also sends adjustment commands to the motors of each magnetic ball.

4. The hybrid control method for permanent magnet arrays according to claim 3, characterized in that: Risk indicators The calculation formula is as follows: ； In the above formula, Indicates and J Relevant condition number normalization indices; Indicates joint velocity; Indicates the rate of change of the magnetic field; The condition number representing the current state; Indicates used to The reference condition number normalized to a reference scale; H B This represents the derivative of the magnetic field with respect to the joint; Indicates a damped pseudo-reverse; Indicates the upper limit of joint velocity; Represents the expected state of the magnetic field during the process. U ref The first derivative; B Indicates magnetic field strength; t Indicates time; Indicates the upper limit of the slope of the magnetic field; express q The first derivative of , i.e., joint velocity.

5. The hybrid control method for permanent magnet arrays according to claim 4, characterized in that, path The search method is as follows: (1) Using each degree of freedom as a node and the influence of each degree of freedom on other degrees of freedom during adjustment as an edge, a directed acyclic graph is formed. G ; (2) Only incentivize any one i When the current value of one degree of freedom increases to the expected value, the other degrees of freedom... j The normalized value of the maximum state change of each degree of freedom is used as the crosstalk index. Construct the cross-activation matrix C: (3) Combining crosstalk index Setting diagram G any directed edge in i → j edge weight g ij The following is a sequence cost graph: ； In the above formula, SatRatio ( q ) indicates and q The reciprocal measure of the relevant minimum attitude limit margin; RateRatio This represents the ratio of the rate of change of the magnetic field to the upper limit of the rate of change of the gradient; w 1. w 2. w 3. w 4 represents the weights of the four indicators; (4) Based on the edge weights of the sequential cost graph g ij Find the acyclic path π that minimizes the total cost of the path by using dynamic programming or integer programming, which allows traversal of all nodes.

6. The hybrid control method for permanent magnet arrays according to claim 5, characterized in that: In secondary planning, action punishment P The expression is as follows: ； The five penalty terms included in the above formula are, in order: tracking penalty, motion amplitude penalty, spatial over-limit penalty, magnetic field climbing penalty, and gradient climbing penalty; W U , W q , W N , W B , W G These are the weight matrices for each penalty term; These represent the influence weights of the last four penalty items; I Represents the identity matrix; This represents the derivative of the gradient with respect to the joint.

7. The hybrid control method for permanent magnet arrays according to claim 6, characterized in that: In quadratic programming, the physical constraints that need to be satisfied include: 、 、 、 、 ； and q It satisfies the positional constraints of the joints and the spatial constraints to avoid collisions; in, They represent q The second and third derivatives, namely joint acceleration and the rate of change of joint acceleration; q velmax , q accmax and q jerkmax Representing joint velocity, joint acceleration, and jerk The upper limit; This represents the upper limit of the gradient slope.

8. The hybrid control method for permanent magnet arrays according to claim 7, characterized in that: If an optimal solution that satisfies the constraints cannot be obtained in any quadratic programming iteration based on parallel mode, the process returns to sequential mode to solve the problem again. When a constrained optimization result cannot be obtained in any sequential pattern-based quadratic programming iteration, the control points for the corresponding interval of the motion trajectory are... s k Encrypt or preset action time Extend the solution and then solve it again. The time dilation factor used to extend the action time The rate of change of automatically contracts according to the tightest constraint and satisfies the following equation: ； In the above formula, It is a conservative coefficient. ; The difference of 8 degrees of freedom between the current state and the target state.

9. The hybrid control method for permanent magnet arrays according to claim 3, characterized in that: A neural network-based magnetic field simulation proxy model is pre-built and trained to simulate the magnetic control mapping of the magnetic field state in the target area under different array postures. Construct an optimization mode proxy model based on graph neural network, which uses each degree of freedom as a node, combines the crosstalk of each degree of freedom to other degrees of freedom when adjusting each degree of freedom to set edge weights and construct a sequential cost graph; in the sequential cost graph, search for the acyclic path π that can traverse all nodes and minimize the total path cost, and use the order of nodes in π as the optimization order of each degree of freedom in the sequential optimization mode. Construct a reinforcement learning-based planning and decision-making agent model to minimize action penalties. P To optimize the objective, following the specified optimization mode, output the joint velocities of each magnetic sphere that can follow the desired magnetic field state and satisfy all physical constraints. ; and according to the preset action time For joint velocity Perform Euler integration to calculate the spatial attitude q of the magnetic sphere array after the control action is performed. t+1 : ; and then combine and q t+1 Generate optimal adjustment action a t ; A federated model comprising a magnetic field simulation agent model, an optimization mode agent model, and a planning decision agent model is used to implement the control method for the permanent magnet array as described in any one of claims 3-8.

10. A magnetic navigation system, characterized in that: It includes two mirror-mounted magnetic control units and a frame capable of driving the lifting and lowering movement of the permanent magnet array; the magnetic control unit comprises a permanent magnet array for a magnetic navigation system as described in claim 1 or 2; The magnetic navigation system further includes a controller; the controller employs a hybrid control method for permanent magnet arrays as described in any one of claims 3-9 to coordinately control two mirror-arranged permanent magnet arrays in the magnetic navigation system, issuing commands to the two motors of each magnetic ball to ensure a preset path within the target area. S upper control points s k The magnetic field strength follows the desired direction to manipulate the magnetically controlled target to move along a preset path.

Images