Spherical motor and position estimation method

Abstract

This invention discloses a spherical motor and a position estimation method, relating to the field of motor technology. The spherical motor includes a stator mechanism and a rotor mechanism. The stator mechanism includes a stator body, an electromagnetic coil assembly disposed on the stator body, and a Hall sensor. The stator body has a spherical cavity. The rotor mechanism includes a rotor body, a connecting block, an actuating magnet assembly, and a positioning magnet. The rotor body is movably disposed in the spherical cavity via a spherical pair. The connecting block is disposed on the rotor body and extends out of the stator body. The actuating magnet assembly is disposed on the rotor body and corresponds to the electromagnetic coil assembly. The positioning magnet is disposed on the rotor body. The Hall sensor interacts with the positioning magnet to acquire the magnetic induction intensity. The technical solution provided by this invention can solve the technical problems of low control accuracy and poor integration in spherical motors.

Description

Technical Field

[0001] This invention relates to the field of motor technology, and in particular to a spherical motor and a position estimation method. Background Art

[0002] A spherical motor, also known as a spherical servo motor, is a special type of motor. Spherical motors allow the rotor to rotate in three-dimensional space for precise control in multiple degrees of freedom. However, traditional spherical motors suffer from low control accuracy and poor integration.

[0003] Therefore, it is necessary to provide a new spherical motor and position estimation method to solve the above-mentioned technical problems. Summary of the Invention

[0004] The main objective of this invention is to provide a spherical motor and a position estimation method, aiming to solve the technical problems of low control accuracy and poor integration of spherical motors.

[0005] To achieve the above objectives, the present invention provides a spherical motor, comprising:

[0006] A stator mechanism, comprising a stator body, an electromagnetic coil assembly disposed on the stator body, and a Hall sensor, wherein the stator body is provided with a spherical cavity;

[0007] The rotor mechanism includes a rotor body, a connecting block, an actuating magnet assembly, and a positioning magnet. The rotor body is movably disposed in the spherical cavity via a spherical pair. The connecting block is disposed on the rotor body and extends out of the stator body. The actuating magnet assembly is disposed on the rotor body, and the positioning magnet is disposed on the rotor body. An electromagnetic coil assembly is used to drive the actuating magnet assembly to move, thereby driving the rotor body to move. A Hall sensor interacts with the positioning magnet and acquires the magnetic induction intensity.

[0008] In one embodiment, the stator body includes a stator base and a stator cover. The stator base is hemispherical, and the stator cover is disposed on the stator base and surrounds the stator base to form the spherical cavity. The electromagnetic coil group is disposed on the stator base, and the Hall sensor is disposed on the stator cover.

[0009] In one embodiment, the stator cover is provided with a mounting boss extending in the circumferential direction of the stator base, the mounting boss is provided with a circuit board, and the circuit board is provided with a plurality of Hall sensors evenly spaced along the circumferential direction of the stator base.

[0010] In one embodiment, the stator base has a hemispherical mounting surface, and a plurality of mounting grooves are evenly arranged on the mounting surface, the mounting grooves being frustoconical in shape;

[0011] The electromagnetic coil assembly includes multiple coil units, and the number of coil units is equal to the number of mounting slots. The multiple coil units are arranged one-to-one in the multiple mounting slots.

[0012] In one embodiment, the rotor body has a first hemisphere and a second hemisphere, the first hemisphere being disposed toward the stator base and the second hemisphere being disposed toward the stator cover, and the actuating magnet assembly being disposed on the first hemisphere and the positioning magnet being disposed on the second hemisphere.

[0013] In one embodiment, there are multiple actuating magnet assemblies, which are spaced apart along the central axis of the rotor body.

[0014] Each of the actuating magnet groups includes multiple permanent magnets, which are evenly spaced along the circumferential direction of the rotor body, and the polarities of the multiple permanent magnets are arranged alternately.

[0015] The number of positioning magnets is multiple, and the multiple positioning magnets are evenly spaced along the circumferential direction of the rotor body.

[0016] This invention also proposes a position estimation method applied to the aforementioned spherical motor, wherein the spherical motor is mounted to a fixed base, a detection rod with visual markers is mounted on the connecting block, and a camera is mounted on one side of the spherical motor, the camera being used to acquire position data of the visual markers; the position estimation method includes:

[0017] The magnetic field strength of the Hall sensor is obtained and input into the training model;

[0018] Obtain the position information of the detection rod;

[0019] The steps to obtain the training model include:

[0020] The detection data of the Hall sensor and the position data of the visual marker are acquired. The detection data of the Hall sensor includes the magnetic induction intensity of the Hall sensor at different time points, and the position data of the visual marker includes the position information of the detection rod at different time points. Data matching is performed based on UNIX timestamps to obtain data pairs including the magnetic induction intensity of the Hall sensor and the position information of the visual marker. Each data pair forms a data set.

[0021] Based on the dataset, a mapping relationship is constructed between the magnetic induction intensity of the Hall sensor and the position information of the visual marker, and a neural network is constructed based on the mapping relationship to generate a training model.

[0022] In one embodiment, after acquiring data pairs including the magnetic induction intensity of the Hall sensor and the position information of the visual marker, and before each data pair is combined into a data set, the method further includes:

[0023] The difference between the magnetic induction intensity of the Hall sensor and the UNIX timestamp of the position information of the visual marker is compared with the value of a preset threshold.

[0024] If the difference between the UNIX timestamp of the Hall sensor's magnetic field strength and the position information of the detection rod in the data pair is less than a preset threshold, then the data pair is output; if the difference between the UNIX timestamp of the Hall sensor's magnetic field strength and the position information of the detection rod in the data pair is greater than a preset threshold, then the data pair is discarded.

[0025] In one embodiment, the steps of constructing a mapping relationship between the magnetic induction intensity of the Hall sensor and the position information of the visual marker based on the data set, and constructing a neural network based on the mapping relationship to generate a training model include: selecting a portion of the data set, constructing a mapping relationship between the magnetic induction intensity of the Hall sensor and the position information of the visual marker based on the portion of the data set, and constructing a neural network based on the mapping relationship to generate a training model;

[0026] The training model is acquired and imported into the computer, and the mapping relationship between the magnetic induction intensity of the Hall sensor and the position information of the visual marker is output. When the magnetic induction intensity of the Hall sensor is the same, the difference between the position information of the visual marker output by the computer and the position information of the visual marker in another part of the data set is compared with the first preset error value.

[0027] If the difference between the position information of the visual marker output by the computer and the position information of the visual marker in another part of the dataset is less than the first preset error value, then the training model meets the requirements; if the difference between the position information of the visual marker output by the computer and the position information of the visual marker in another part of the dataset is greater than the first preset error value, then the training model does not meet the requirements, the neural network is optimized, a training model is generated, and the steps of obtaining the training model and importing it into the computer are executed.

[0028] The training model is obtained and imported into the controller of the spherical motor. The difference between the position information of the visual marker output by the controller and the position information of the visual marker obtained by the camera at the same time point is compared with the second preset error value.

[0029] If the difference between the position information of the visual marker output by the controller and the position information of the visual marker acquired by the camera is less than the second preset error value, then the training model meets the requirements; if the difference between the position information of the visual marker output by the controller and the position information of the visual marker acquired by the camera is greater than the second preset error value, then the training model does not meet the requirements, the neural network is optimized, a training model is generated, and the steps of acquiring the training model and importing it into the computer are executed.

[0030] In one embodiment, after the step of obtaining the trained model, the method further includes:

[0031] The training model is obtained and imported into the controller of the spherical motor. The detection rod is removed, and the spherical motor is installed on a mechanical device capable of three-degree-of-freedom control. The controller of the mechanical device is electrically connected to the controller of the spherical motor.

[0032] The technical solution of this invention achieves relative movement between the rotor body and the stator body through a spherical pair. Furthermore, it acquires magnetic induction intensity through the interaction of a Hall sensor and a positioning magnet, thereby calculating the rotor body's position information. This reduces friction in the spherical motor, improves its control accuracy, and enhances its integration. In this embodiment, the rotor body is movably disposed within the spherical cavity of the stator body via the spherical pair, reducing friction during motor operation and improving its operational accuracy. The Hall sensor interacts with the positioning magnet and transmits the acquired magnetic induction intensity to the spherical motor's controller. The controller, using a training model incorporating a neural network, converts the magnetic induction intensity acquired by the Hall sensor into position information of the detection rod, thus calculating the rotor body's position information. Since the controller acquires the rotor body's position information without contacting it, it reduces friction during motor operation and improves the motor's operational accuracy. The actuating magnet assembly can interact with the electromagnetic coil assembly to drive the rotor body to move. Specifically, when the electromagnetic coil assembly is energized, the electromagnetic coil assembly generates a magnetic field. By changing the magnitude and direction of the current, the magnetic field generated by the electromagnetic coil assembly will change, which will then drive the actuating magnet assembly to move, thereby driving the rotor body to move. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0034] Figure 1 This is a schematic diagram of the structure of the spherical motor in the embodiments provided by the present invention;

[0035] Figure 2 This is a schematic diagram of the stator mechanism in an embodiment of the present invention;

[0036] Figure 3 for Figure 2 Another perspective illustration;

[0037] Figure 4 This is a schematic diagram of the rotor mechanism in an embodiment of the present invention;

[0038] Figure 5 for Figure 4 Another perspective illustration;

[0039] Figure 6 This is a schematic diagram of the installation of the spherical motor when obtaining the training model in an embodiment of the present invention;

[0040] Figure 7 A flowchart of the location estimation method provided in the embodiments of the present invention;

[0041] Figure 8 A flowchart illustrating the acquisition of a training model in an embodiment of the present invention.

[0042] Explanation of icon numbers:

[0043] 100. Stator mechanism; 110. Stator body; 111. Spherical cavity; 112. Stator base; 1121. Mounting surface; 1122. Mounting groove; 113. Stator cover; 1131. Mounting boss; 200. Rotor mechanism; 210. Rotor body; 211. First hemisphere; 212. Second hemisphere; 220. Connecting block; 230. Actuating magnet assembly; 231. Permanent magnet; 240. Positioning magnet; 310. Detection rod; 320. Fixing base; 330. Camera.

[0044] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0046] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0047] Furthermore, if the embodiments of the present invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, if "and / or" or "and / or" appears throughout the text, its meaning includes three parallel solutions. For example, "A and / or B" includes solution A, solution B, or a solution where both A and B are satisfied simultaneously.

[0048] Furthermore, the technical solutions of the various embodiments of the present invention can be combined with each other, but only if they are feasible for those skilled in the art. If the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0049] A spherical motor, also known as a ball motor or spherical servo motor, allows the rotor to rotate in three-dimensional space for precise control in multiple degrees of freedom. A spherical motor consists of a rotor and a stator. In practical applications, researchers have found that traditional spherical motors mostly use ball bearings to support the rotor, which increases friction between the rotor and stator, affecting the motor's operating accuracy. Furthermore, during operation, the rotor's position needs to be sensed in real time. Traditional spherical motors mostly use a contact-type structure for position sensing, which requires contact with the rotor, increasing friction and affecting the motor's operating accuracy.

[0050] This invention proposes a spherical motor and a position estimation method, aiming to solve the technical problems of low control accuracy and poor integration of spherical motors.

[0051] Please see Figure 1 , Figure 2 and Figure 4In one embodiment of the present invention, the spherical motor includes a stator mechanism 100 and a rotor mechanism 200. The stator mechanism 100 includes a stator body 110, an electromagnetic coil assembly disposed on the stator body 110, and a Hall sensor. The stator body 110 is provided with a spherical cavity 111. The rotor mechanism 200 includes a rotor body 210, a connecting block 220, an actuating magnet assembly 230, and a positioning magnet 240. The rotor body 210 is movably disposed in the spherical cavity 111 via a spherical pair. The connecting block 220 is disposed on the rotor body 210 and extends out of the stator body 110. The actuating magnet assembly 230 is disposed on the rotor body 210, and the positioning magnet 240 is disposed on the rotor body 210. The electromagnetic coil assembly is used to drive the actuating magnet assembly 230 to move, thereby driving the rotor body 210 to move. The Hall sensor interacts with the positioning magnet and obtains the magnetic induction intensity.

[0052] The technical solution of this invention achieves relative movement between the rotor body 210 and the stator body 110 through a spherical pair. Furthermore, it obtains magnetic induction intensity through the interaction of a Hall sensor and a positioning magnet, thereby calculating the position information of the rotor body 210. This reduces friction in the spherical motor, improves its control accuracy, and enhances its integration. In this embodiment, the rotor body 210 is movably disposed within the spherical cavity 111 of the stator body 110 via the spherical pair, reducing friction during motor operation and improving its operational accuracy. The Hall sensor interacts with the positioning magnet 240 and transmits the acquired magnetic induction intensity to the spherical motor's controller. The controller then uses a training model incorporating a neural network to convert the magnetic induction intensity acquired by the Hall sensor into position information of the detection rod 310, thereby calculating the position information of the rotor body 210. Since the controller acquires the position information of the rotor body 210 without contacting it, it reduces friction during motor operation and improves the motor's operational accuracy. The actuating magnet assembly 230 can interact with the electromagnetic coil assembly to drive the rotor body 210 to move. Specifically, when the electromagnetic coil assembly is energized, the electromagnetic coil assembly generates a magnetic field. By changing the magnitude and direction of the current, the magnetic field generated by the electromagnetic coil assembly changes, which then drives the actuating magnet assembly 230 to move, thereby driving the rotor body 210 to move.

[0053] It should be noted that the use of a spherical pair to achieve the relative movement of the rotor body 210 and the stator body 110 reduces friction during the operation of the spherical motor because: the spherical pair ensures that the rotor body 210 and the stator body 110 are in spherical contact, and the spherical contact creates an arc-shaped contact area, which helps to disperse contact stress and reduce friction between the rotor body 210 and the stator body 110. At the same time, using a spherical pair connection also improves the load-bearing capacity of the spherical motor, simplifies its structure, and reduces manufacturing costs.

[0054] Please see Figure 2 and Figure 3 In one embodiment of the present invention, the stator body 110 includes a stator base 112 and a stator cover 113. The stator base 112 is hemispherical, and the stator cover 113 is disposed on the stator base 112 and surrounds the stator base 112 to form a spherical cavity 111. An electromagnetic coil assembly is disposed on the stator base 112, and a Hall sensor is disposed on the stator cover 113. In this embodiment, designing the stator base 112 as a hemispherical shape can reduce the volume of the spherical motor, facilitating its installation. In a specific embodiment, the stator base 112 has an extension plate along its circumference, and the extension plate is connected to the stator cover 113 by bolts. The stator base 112 is a plastic or metal part made using 3D printing technology. The metal part has better thermal conductivity, and using a metal stator base 112 can improve the stability of the spherical motor during operation.

[0055] In one embodiment of the present invention, the stator cover 113 is provided with a mounting boss 1131 extending in the circumferential direction of the stator base 112. A circuit board is mounted on the mounting boss 1131, and multiple Hall sensors are evenly spaced along the circumferential direction of the stator base 112 on the circuit board. In this embodiment, the Hall sensors interact with the positioning magnet 240 and transmit the acquired magnetic induction intensity to the controller of the spherical motor. The controller can then convert the magnetic induction intensity acquired by the Hall sensors into position information of the detection rod 310 through a training model containing a neural network, thereby calculating the position information of the rotor body 210. Using a Hall sensor in conjunction with the positioning magnet 240 to acquire the position information of the rotor body 210 allows for more accurate determination of the rotor body 210's position, achieving high-precision position recognition. Furthermore, the Hall sensor does not need to contact the rotor body 210 when acquiring the position of the positioning magnet 240, reducing friction during spherical motor operation and improving the spherical motor's operating accuracy. Simultaneously, the Hall sensor also has the advantages of small size and low cost, enabling a reduction in the size of the spherical motor and lowering its manufacturing cost. In one specific embodiment, the circuit board is circular or C-shaped, and the number of Hall sensors is four. The four Hall sensors are evenly spaced on the circuit board along the circumferential direction of the stator base 112.

[0056] In one embodiment of the present invention, the stator base 112 has a hemispherical mounting surface 1121, on which a plurality of mounting grooves 1122 are evenly arranged, and the mounting grooves 1122 are frustum-shaped. The electromagnetic coil assembly includes a plurality of coil units, and the number of coil units and mounting grooves 1122 are equal, with the plurality of coil units correspondingly disposed in the plurality of mounting grooves 1122. In this embodiment, each coil unit is frustum-shaped. By opening a plurality of frustum-shaped mounting grooves 1122 on the stator base 112 and distributing the plurality of coil units correspondingly in the plurality of mounting grooves 1122, the arrangement density of the coil units on the mounting surface 1121 of the stator base 112 can be increased, thereby increasing the output torque of the spherical motor. The reason is that installing the frustum-shaped coil unit into the frustum-shaped mounting slot 1122 effectively avoids interference when the coil unit is installed into the stator base 112. If both the mounting slot 1122 and the coil unit are cylindrical, since the diameter of the stator base 112 gradually decreases from the outside to the inside, interference is highly likely to occur when the coil unit is installed into the stator base 112 at the ends of two adjacent coil units inserted into the stator base 112. In a specific embodiment, there are eight mounting slots 1122 and eight coil units. Among the eight mounting slots 1122, two mounting slots 1122 are located on the side of the stator base 112 away from the stator cover 113, and the other six mounting slots 1122 are evenly spaced outside the two mounting slots 1122. Furthermore, four of the eight mounting slots 1122 are arranged in an arc shape. In this embodiment, the controller of the spherical motor includes a drive board, which is connected to the coil unit via wires or the like. By running the drive control program of the spherical motor, the position of the rotor body 210 of the spherical motor can be adjusted.

[0057] Please see Figure 4 and Figure 5In one embodiment of the present invention, the rotor body 210 has a first hemisphere 211 and a second hemisphere 212. The first hemisphere 211 faces the stator base 112, and the second hemisphere 212 faces the stator cover 113. An actuating magnet assembly 230 is disposed on the first hemisphere 211, and a positioning magnet 240 is disposed on the second hemisphere 212. In this embodiment, by disposing the actuating magnet assembly 230 and the positioning magnet 240 on opposite hemispheres of the rotor body 210, the distance between the actuating magnet assembly 230 and the positioning magnet 240 is increased, thereby increasing the distance between the actuating magnet assembly 230 and the Hall sensor. This reduces the influence of the actuating magnet assembly 230 on the Hall sensor, resulting in more accurate detection data from the Hall sensor. Correspondingly, a mounting boss 1131 is provided on the stator cover 113, and the circuit board with the Hall sensor is mounted on the mounting boss 1131. This also increases the distance between the actuating magnet assembly 230 and the Hall sensor, making the detection data from the Hall sensor more accurate. In one specific embodiment, there are multiple positioning magnets 240, which are evenly spaced along the circumferential direction of the rotor body 210. In a more specific embodiment, there are two positioning magnets 240, which are symmetrically arranged and have opposite polarities.

[0058] In one embodiment of the present invention, there are multiple actuating magnet groups 230, which are spaced apart along the central axis of the rotor body 210. Each actuating magnet group 230 includes multiple permanent magnets 231, which are evenly spaced along the circumferential direction of the rotor body 210, and the polarities of the permanent magnets 231 are arranged alternately. In this embodiment, arranging the polarities of the permanent magnets 231 alternately generates a more uniform and continuous magnetic field, thereby generating a larger torque on the rotor body 210, increasing the motion angle of the spherical motor and improving its output torque. Simultaneously, it also provides more flexible magnetic field control, enhancing the dynamic performance of the spherical motor. In a specific embodiment, there are two actuating magnet groups 230, each including four permanent magnets 231, meaning that the rotor body of the spherical motor has eight permanent magnets 231 installed.

[0059] Please see Figure 7 The present invention also proposes a position estimation method for the above-mentioned spherical motor, the position estimation method comprising:

[0060] S100: Obtain the magnetic induction intensity of the Hall sensor and input it into the training model;

[0061] S200: Obtain the position information of the detection rod 310; subsequently, the controller can obtain the relative attitude of the rotor body 210 of the spherical motor based on the obtained position information of the detection rod 310.

[0062] Please see Figure 6 When using this position estimation method, a trained model containing a neural network needs to be acquired first. Before this, the spherical motor needs to be installed on the mounting base 320, a detection rod 310 with visual markers needs to be installed on the connecting block 220, and a camera 330 for acquiring position data of the visual markers needs to be installed on one side of the spherical motor. The detection rod 310 with visual markers can be a detection rod 310 with a CylinderTag code. In this embodiment, the connecting block 220 has a slot, and the detection rod 310 has a locking block. The detection rod 310 is installed on the connecting block 220 by locking the locking block into the slot. It should be noted that the CylinderTag code is a visual marker specifically designed for cylindrical objects, aiming to achieve high-precision attitude estimation. It encodes information in the zero-curvature direction of the surface by utilizing the cross ratio in projection invariance, thereby achieving accurate tracking and positioning of cylindrical objects.

[0063] Please see Figure 8 The steps to obtain the training model include:

[0064] The process involves acquiring detection data from a Hall sensor and position data from visual markers. The Hall sensor detection data includes the magnetic field strength of the Hall sensor at different time points, while the visual marker position data includes the position information of the visual markers at different time points. Data matching is performed based on UNIX timestamps to obtain data pairs containing the Hall sensor's magnetic field strength and the visual marker's position information; these data pairs form a dataset. Based on this dataset, a mapping relationship between the Hall sensor's magnetic field strength and the visual marker's position information is constructed, and a neural network is built based on this mapping relationship to generate a training model. The position information of the visual markers can be quaternion data from CylinderTag codes.

[0065] Specifically, the controller of the spherical motor can acquire the magnetic induction intensity of the Hall sensor and send the data to the computer in real time via serial communication to obtain the detection data of the Hall sensor. Simultaneously, the shooting data from the camera 330 can be transmitted to the computer. The computer processes the video data captured by the camera 330 and uses the CylinderTag attitude estimation algorithm to calculate the attitude of the detection rod 310 in real time (when recording the attitude of the detection rod 310, an attitude zero point can be manually set, and the relative attitude between other positions and the attitude zero point can be calculated) to obtain the position data of the visual marker. Subsequently, data matching is performed based on the UNIX timestamp (also known as the POSIX timestamp or Epoch time, which refers to the number of seconds elapsed from January 1, 1970 to the current time), selecting data pairs that are close in time and forming a data set. Based on the data set, a mapping relationship between the magnetic induction intensity of the Hall sensor and the position information of the visual marker is constructed. The computer uses the matched data set to train a neural network through the backpropagation algorithm to generate a training model. During training, the model learns how to predict the position information of the corresponding visual marker based on the magnetic induction intensity of the input Hall sensor, and minimizes the difference between the predicted output and the actual output by adjusting the internal weights and biases. In an embodiment of the invention, the input layer consists of 12 magnetic induction intensities from the Hall sensor, and the output layer consists of 4 quaternion data points of the visual marker.

[0066] In one embodiment of the present invention, after acquiring data pairs including the magnetic induction intensity of the Hall sensor and the position information of the visual marker, before assembling the data sets from the data pairs, the method further includes: comparing the UNIX timestamp difference between the magnetic induction intensity of the Hall sensor and the position information of the visual marker with a preset threshold; if the UNIX timestamp difference between the magnetic induction intensity of the Hall sensor and the position information of the detection rod 310 of the data pair is less than the preset threshold, the data pair is output; if the UNIX timestamp difference between the magnetic induction intensity of the Hall sensor and the position information of the detection rod 310 of the data pair is greater than the preset threshold, the data pair is discarded. Retaining the portion where the UNIX timestamp difference between the magnetic induction intensity of the Hall sensor and the position information of the visual marker is less than the preset threshold ensures that the magnetic induction intensity of the Hall sensor and the position information of the visual marker in the data pair were acquired at similar time points, thereby more accurately analyzing the relationship between the two. In this embodiment, the preset threshold can be 5ms.

[0067] In one embodiment of the present invention, the steps of constructing a mapping relationship between the magnetic induction intensity of the Hall sensor and the position information of the visual marker based on a dataset, and constructing a neural network based on the mapping relationship to generate a training model include: selecting a portion of the dataset, constructing a mapping relationship between the magnetic induction intensity of the Hall sensor and the position information of the visual marker based on the portion of the dataset, constructing a neural network based on the mapping relationship, and generating a training model.

[0068] The training model is acquired and imported into the computer, and the mapping relationship between the magnetic induction intensity of the Hall sensor and the position information of the visual marker is output. When the magnetic induction intensity of the Hall sensor is the same, the difference between the position information of the visual marker output by the computer and the position information of the visual marker in another dataset is compared to a first preset error value. If the difference between the position information of the visual marker output by the computer and the position information of the visual marker in another dataset is less than the first preset error value, the training model meets the requirements. If the difference between the position information of the visual marker output by the computer and the position information of the visual marker in another dataset is greater than the first preset error value, the training model does not meet the requirements. The neural network is then optimized, a training model is generated, and the steps of acquiring the training model and importing it into the computer are executed.

[0069] The training model is acquired and imported into the controller of the spherical motor. The difference between the position information of the visual marker output by the controller and the position information of the visual marker acquired by the camera 330 at the same time point is compared with a second preset error value. If the difference between the position information of the visual marker output by the controller and the position information of the visual marker acquired by the camera 330 is less than the second preset error value, the training model meets the requirements. If the difference between the position information of the visual marker output by the controller and the position information of the visual marker acquired by the camera 330 is greater than the second preset error value, the training model does not meet the requirements. The neural network is then optimized, a training model is generated, and the steps of acquiring the training model and importing it into the computer are executed.

[0070] Specifically, when generating the training model, 70%-80% of the dataset is selected. After the training model is generated, its initial performance can be verified based on the remaining 20%-30% of the dataset. Specifically, the training model is acquired and imported into a computer. Then, the difference between the positional information of the visual markers output by the computer and the positional information of the visual markers in the remaining 20%-30% of the dataset, when the magnetic induction intensity of the Hall sensor is the same, is compared to a first preset error value to verify the initial performance of the training model. If the difference is less than the first preset error value, the training model meets the requirements; if the difference is greater than the first preset error value, the training model does not meet the requirements. Subsequently, the neural network is further adjusted (e.g., increasing or decreasing the number of layers, adjusting the number of neurons, etc.) to generate a new training model. The steps of acquiring the training model and importing it into the computer are then repeated until the training model meets the requirements.

[0071] Similarly, after verifying the initial performance of the trained model on the computer, the trained model can be acquired and imported into the controller of the spherical motor to evaluate its performance in the hardware environment. Specifically, the trained model is imported into the controller of the spherical motor, and the difference between the position information of the visual marker output by the controller and the position information of the visual marker acquired by the camera 330 at the same time point is compared with a second preset error value. If the difference is less than the second preset error value, it means that the trained model meets the requirements, that is, the trained model performs well in the hardware environment; if the difference is greater than the second preset error value, it means that the trained model does not meet the requirements, that is, the trained model performs poorly in the hardware environment. Subsequently, the neural network is further adjusted (e.g., increasing or decreasing the number of layers, adjusting the number of neurons, etc.) to generate a new trained model, and then the steps of acquiring the trained model and importing it into the computer are repeated until the trained model meets the requirements.

[0072] Following the step of acquiring the training model, the process further includes: acquiring the training model and importing it into the controller of the spherical motor, removing the detection rod, and installing the spherical motor onto a mechanical device capable of three-degree-of-freedom control. The controller of the mechanical device is electrically connected to the controller of the spherical motor. Specifically, after completing the initial calibration of the training model, the spherical motor is installed onto the mechanical device capable of three-degree-of-freedom control. By acquiring the magnetic induction intensity of the Hall sensor, the rotor's position information can be calculated using the training model, which includes a neural network, thus achieving autonomous position perception without the detection rod. In a specific embodiment, the mechanical device capable of three-degree-of-freedom control can be a cross-medium robot. The cross-medium robot includes a fuselage, wings, and a controller. The spherical motor is installed on the fuselage, and the rotor body of the spherical motor is connected to the wings. The spherical motor is electrically connected to the controller. After completing the initial calibration of the training model, the spherical motor can be installed on the cross-medium robot. The spherical motor drives the wings to swing. The controller of the cross-medium robot is electrically connected to the spherical motor, allowing the controller to acquire the position information of the rotor body in real time through the spherical motor, thereby obtaining the relative attitude of the wings.

[0073] In this embodiment, to prevent overfitting during training, an early stopping mechanism is introduced to ensure the training model has good generalization ability. Overfitting refers to a situation where the model performs well on training data but poorly on unseen data.

[0074] The above description is merely an exemplary embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural transformations made using the contents of the present invention specification and drawings under the technical concept of the present invention, or direct / indirect applications in other related technical fields, are included within the patent protection scope of the present invention.

Claims

1. A spherical motor, characterized in that, include: A stator mechanism, comprising a stator body, an electromagnetic coil assembly disposed on the stator body, and a Hall sensor, wherein the stator body is provided with a spherical cavity; The rotor mechanism includes a rotor body, a connecting block, an actuating magnet assembly, and a positioning magnet. The rotor body is movably disposed in the spherical cavity via a spherical joint. The connecting block is disposed on the rotor body and extends out of the stator body. The actuating magnet assembly is disposed on the rotor body, and the positioning magnet is disposed on the rotor body. An electromagnetic coil assembly is used to drive the actuating magnet assembly to move, thereby driving the rotor body to move. A Hall sensor interacts with the positioning magnet and acquires the magnetic induction intensity. The stator body includes a stator base and a stator cover. The stator base is hemispherical, and the stator cover is disposed on the stator base and surrounds the stator base to form the spherical cavity. The electromagnetic coil group is disposed on the stator base, and the Hall sensor is disposed on the stator cover. The stator cover is provided with a mounting boss extending in the circumferential direction of the stator base. The mounting boss is provided with a circuit board, and the circuit board is provided with a plurality of Hall sensors evenly spaced in the circumferential direction of the stator base. The rotor body has a first hemisphere and a second hemisphere. The first hemisphere faces the stator base, and the second hemisphere faces the stator cover. The actuating magnet assembly is disposed on the first hemisphere, and the positioning magnet is disposed on the second hemisphere. The number of actuating magnet groups is multiple, and the multiple actuating magnet groups are spaced apart along the central axis of the rotor body; Each of the actuating magnet groups includes multiple permanent magnets, which are evenly spaced along the circumferential direction of the rotor body, and the polarities of the multiple permanent magnets are arranged alternately. The number of positioning magnets is multiple, and the multiple positioning magnets are evenly spaced along the circumferential direction of the rotor body.

2. The spherical motor as described in claim 1, characterized in that, The stator base has a hemispherical mounting surface, and multiple mounting grooves are evenly arranged on the mounting surface, the mounting grooves being frustoconical in shape; The electromagnetic coil assembly includes multiple coil units, and the number of coil units is equal to the number of mounting slots. The multiple coil units are arranged one-to-one in the multiple mounting slots.

3. A position estimation method, applied to a spherical motor as described in any one of claims 1 to 2, characterized in that, The spherical motor is mounted to the fixed base, a detection rod with visual markers is mounted on the connecting block, and a camera is mounted on one side of the spherical motor. The camera is used to acquire the position data of the visual markers. The location estimation method includes: The magnetic field strength of the Hall sensor is obtained and input into the training model; Obtain the position information of the detection rod; The steps to obtain the training model include: The detection data of the Hall sensor and the position data of the visual marker are acquired. The detection data of the Hall sensor includes the magnetic induction intensity of the Hall sensor at different time points, and the position data of the visual marker includes the position information of the detection rod at different time points. Data matching is performed based on UNIX timestamps to obtain data pairs including the magnetic induction intensity of the Hall sensor and the position information of the visual marker. Each data pair forms a data set. Based on the dataset, a mapping relationship is constructed between the magnetic induction intensity of the Hall sensor and the position information of the visual marker, and a neural network is constructed based on the mapping relationship to generate a training model.

4. The location estimation method as described in claim 3, characterized in that, After acquiring data pairs including the magnetic induction intensity of the Hall sensor and the position information of the visual marker, and before the data pairs are combined into a data set, the process further includes: The difference between the magnetic induction intensity of the Hall sensor and the UNIX timestamp of the position information of the visual marker is compared with the value of a preset threshold. If the difference between the UNIX timestamp of the Hall sensor's magnetic field strength and the position information of the detection rod in the data pair is less than a preset threshold, then the data pair is output; if the difference between the UNIX timestamp of the Hall sensor's magnetic field strength and the position information of the detection rod in the data pair is greater than a preset threshold, then the data pair is discarded.

5. The location estimation method as described in claim 3, characterized in that, The steps of constructing a mapping relationship between the magnetic induction intensity of the Hall sensor and the position information of the visual marker based on the data set, and constructing a neural network based on the mapping relationship to generate a training model include: selecting a portion of the data set, constructing a mapping relationship between the magnetic induction intensity of the Hall sensor and the position information of the visual marker based on the portion of the data set, and constructing a neural network based on the mapping relationship to generate a training model; The training model is acquired and imported into the computer, and the mapping relationship between the magnetic induction intensity of the Hall sensor and the position information of the visual marker is output. When the magnetic induction intensity of the Hall sensor is the same, the difference between the position information of the visual marker output by the computer and the position information of the visual marker in another part of the data set is compared with the first preset error value. If the difference between the position information of the visual marker output by the computer and the position information of the visual marker in another part of the dataset is less than the first preset error value, then the training model meets the requirements; if the difference between the position information of the visual marker output by the computer and the position information of the visual marker in another part of the dataset is greater than the first preset error value, then the training model does not meet the requirements, the neural network is optimized, a training model is generated, and the steps of obtaining the training model and importing it into the computer are executed. The training model is obtained and imported into the controller of the spherical motor. The difference between the position information of the visual marker output by the controller and the position information of the visual marker obtained by the camera at the same time point is compared with the second preset error value. If the difference between the position information of the visual marker output by the controller and the position information of the visual marker acquired by the camera is less than the second preset error value, then the training model meets the requirements; if the difference between the position information of the visual marker output by the controller and the position information of the visual marker acquired by the camera is greater than the second preset error value, then the training model does not meet the requirements, the neural network is optimized, a training model is generated, and the steps of acquiring the training model and importing it into the computer are executed.

6. The location estimation method as described in claim 3, characterized in that, After obtaining the trained model, the process further includes: The training model is obtained and imported into the controller of the spherical motor. The detection rod is removed, and the spherical motor is installed on a mechanical device capable of three-degree-of-freedom control. The controller of the mechanical device is electrically connected to the controller of the spherical motor.

Images