Dual-mode neuromorphic magnetic navigation device and navigation method

Abstract

The present application relates to the technical field of magnetic navigation, in particular to a dual-mode neuromorphic magnetic navigation device and a navigation method, which mainly solves the technical problem that the existing magnetic navigation device has low sensitivity, slow response and is difficult to balance long-distance fast orientation and short-distance high-precision positioning. The device comprises a magnetic sensing substrate, a plurality of array-distributed magnetic sensing pixel modules, an array microwave antenna and a microwave source; the magnetic sensing substrate is internally provided with a neural network architecture circuit and a signal conditioning circuit; each magnetic sensing pixel module comprises a regulation and control substrate and at least three fluorescent sensing units, and each fluorescent sensing unit comprises an NV color center, a filter, a photodetector and a laser. The application also discloses a navigation method based on the above device. The application adopts a magnetic compass and a magnetic map dual-mode to work cooperatively, realizes the continuous transition of long-distance rough navigation and short-distance accurate navigation, and has the beneficial effects of high sensitivity, fast response speed and low power consumption.

Description

Technical Field

[0001] This invention relates to the field of magnetic navigation technology, and in particular to a dual-mode neuromorphic magnetic navigation device and navigation method. Background Technology

[0002] With the development of unmanned systems, intelligent aircraft, and autonomous navigation technology in complex environments, how to achieve high-precision navigation of planned routes in environments with weak signals, no satellites, or strong interference has become one of the important research directions in the current navigation field. Traditional satellite navigation is prone to failure in such environments. Magnetic navigation, as an autonomous navigation technology that does not rely on external signal sources, identifies and navigates by acquiring environmental magnetic vector information and spatial magnetic map information. It has high reliability and anti-interference capabilities, and has become a key way to solve this problem.

[0003] However, existing magnetic navigation technologies still have the following shortcomings: At the perception level, existing devices mainly rely on Hall sensors, magnetoresistive sensors, or fiber optic magnetic sensors to detect magnetic fields. Their sensitivity is limited and their response speed is slow in weak magnetic environments, making it difficult to meet the real-time requirements of high-speed dynamic navigation scenarios. At the computing architecture level, magnetic map-based navigation methods usually rely on external processors to perform offline storage and recognition calculations on the collected magnetic field data. The architecture of separating perception and computation not only introduces large system latency but also brings high power consumption and hardware complexity, limiting its application in real-time navigation. At the coordination mechanism level, existing magnetic navigation systems generally lack effective coordination between magnetic vector information and magnetic map information, making it impossible to switch between coarse orientation and precise positioning modes, and making it difficult to meet the dual requirements of long-distance rapid orientation and short-distance high-precision positioning.

[0004] Therefore, there is an urgent need for a dual-mode magnetic navigation device that can simultaneously achieve high sensitivity, ultra-fast response, and the ability to coordinate magnetic vectors and magnetic maps, in order to solve the aforementioned problems. Summary of the Invention

[0005] To overcome the technical shortcomings of existing magnetic navigation devices, such as low sensitivity, slow response, and difficulty in simultaneously meeting the dual requirements of long-distance rapid orientation and short-distance high-precision positioning, this invention proposes a dual-mode neuromorphic magnetic navigation device and navigation method.

[0006] The dual-mode neuromorphic magnetic navigation device provided by this invention includes:

[0007] A magnetic sensing substrate with a built-in neural network architecture circuit and signal conditioning circuit;

[0008] A magnetic sensing pixel module is provided, which is arranged in an array on the magnetic sensing substrate. Each magnetic sensing pixel module includes a control substrate and at least three fluorescence sensing units disposed on the control substrate. The control substrate has a built-in control circuit. Each fluorescence sensing unit includes an NV color center, a filter disposed on the light-emitting side of the NV color center, a photodetector disposed on the light-emitting side of the filter, and a laser arranged towards the NV color center. The at least three fluorescence sensing units are used to acquire signals of NV color centers with different crystal orientations respectively. The laser and the photodetector are both fixed on the control substrate and electrically connected to the control circuit. The control substrate is fixed on the magnetic sensing substrate, and the control circuit is electrically connected to the neural network architecture circuit and the signal conditioning circuit.

[0009] An array microwave antenna is arranged parallel above the magnetic sensing substrate and fixed relative to the magnetic sensing substrate. The array microwave antenna has multiple microwave radiation regions, which are arranged one-to-one with the NV color centers and are used to provide microwave fields for the NV color centers.

[0010] A microwave source, which is connected to the array microwave antenna;

[0011] The neural network architecture circuit is configured to directly perform a weighted summation operation on the output signal of the fluorescence sensing unit in the analog domain and output navigation commands; the signal conditioning circuit is configured to selectively activate some fluorescence sensing units to achieve magnetic compass mode or activate all fluorescence sensing units to achieve magnetic map mode.

[0012] Furthermore, the control substrate is detachably plugged into the magnetic sensing substrate via an electrical interface.

[0013] Furthermore, the control substrate is provided with four fluorescence sensing units arranged in a cross shape.

[0014] Furthermore, multiple magnetic sensing pixel modules are arranged in a square array on the magnetic sensing substrate.

[0015] Furthermore, the laser is a 532nm semiconductor laser, and the filter is a 600nm-800nm ​​bandpass filter.

[0016] Furthermore, each microwave radiation region of the array microwave antenna is provided with an independent feeding network, and each feeding network sets impedance parameters according to the responsivity weight of the corresponding fluorescent sensing unit.

[0017] Furthermore, the power supply network adjusts the responsivity weight of the corresponding fluorescence sensing unit by regulating the microwave frequency and / or power of each microwave radiation region.

[0018] Furthermore, the microwave radiation region is a radiation window opened on the array microwave antenna, and the radiation window corresponds one-to-one with the position of the NV color center.

[0019] Furthermore, the microwave source is fixed relative to the magnetic sensing substrate, and the microwave source is connected to the array microwave antenna via a low-loss radio frequency connection line.

[0020] The dual-mode neuromorphic magnetic navigation method provided by this invention is based on the aforementioned dual-mode neuromorphic magnetic navigation device:

[0021] When the distance between the real-time position and the target position is greater than the preset distance, the magnetic compass mode is used for navigation. In the magnetic compass mode, the signal conditioning circuit selects three fluorescent sensing units with different directional selectivity to work through the control circuit, which correspond to the three-dimensional orthogonal coordinate axes respectively, in order to obtain XYZ three-axis magnetic vector information.

[0022] When the distance between the real-time position and the target position is less than or equal to the preset distance, the magnetic map mode is used for navigation. In the magnetic map mode, the signal conditioning circuit controls all fluorescence sensing units to work through the control circuit to obtain magnetic map information. The neural network architecture circuit directly performs weighted summation on the output signals of all fluorescence sensing units in the analog domain and outputs navigation commands based on the weighted summation result.

[0023] The technical solution provided by this invention has the following advantages compared with the prior art.

[0024] The dual-mode neuromorphic magnetic navigation device provided by this invention uses diamond NV color centers as magnetic sensing units, combined with a side-arranged laser and filter-photodetector integrated structure, to achieve high-sensitivity weak magnetic detection. By integrating the neural network architecture circuit and the magnetic sensing pixel module on the same magnetic sensing substrate, an integrated perception and computing architecture is constructed, which directly performs weighted summation operations in the analog domain, significantly reducing system latency and hardware complexity. Combined with signal conditioning circuitry, selective activation of magnetic compass mode and magnetic map mode is achieved, resulting in a device that combines high sensitivity, low latency, and dual-mode cooperative navigation capabilities.

[0025] The dual-mode neuromorphic magnetic navigation method provided by this invention achieves intelligent switching between modes based on a preset distance. For long-distance navigation, a magnetic compass mode is used, selecting three fluorescent sensing units with different directional selectivity to acquire XYZ three-axis magnetic vector information for rapid orientation. For short-distance navigation, the method switches to magnetograph mode, activating all fluorescent sensing units to acquire magnetograph information. A neural network architecture circuit directly performs a weighted summation operation on the output signals of all fluorescent sensing units in the analog domain, and outputs navigation commands based on the weighted summation result. This method effectively balances the dual requirements of rapid long-distance orientation and high-precision short-distance positioning through its mode-switching mechanism. Attached Figure Description

[0026] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0027] 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, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a top view showing the overall structure of the magnetic navigation device in an embodiment of the present invention;

[0029] Figure 2 This is a side view showing the overall structure of the magnetic navigation device in an embodiment of the present invention.

[0030] Figure 3 A schematic diagram showing the magnetic sensing substrate and its associated structures in an embodiment of the present invention;

[0031] Figure 4 This is a top view of the magnetic sensing pixel module in an embodiment of the present invention;

[0032] Figure 5 This is a side view of the magnetic sensing pixel module in an embodiment of the present invention;

[0033] Figure 6 This diagram illustrates the structure of the array microwave antenna in an embodiment of the present invention.

[0034] Figure 7 The diagram shows the operational logic of the magnetic compass mode in Experiment 1.

[0035] Figure 8 The graph shows the navigation performance test results of the magnetic compass mode in Experiment 1; among them, Figure 8 (a) to Figure 8 (c) shows the comparison results between the triaxial magnetic field and the ideal magnetic field value. Figure 8 (d) represents the statistical results of the vector magnetic field error;

[0036] Figure 9 The operational logic diagram representing the magnetic map mode in Experiment 2;

[0037] Figure 10 The graph shows the navigation performance test results of the magnetograph mode in Experiment 2; among them, Figure 10 (a) to Figure 10 (d) represents the recognition results of the four magnetic field patterns, Figure 10(e) represents the accuracy statistics at different noise levels.

[0038] In the picture:

[0039] 100 Magnetic sensing substrate; 200 Magnetic sensing pixel module; 210 Control substrate; 220 Fluorescent sensing unit; 221 NV color center; 222 Filter; 223 Photodetector; 224 Laser; 230 Electrical interface; 300 Array microwave antenna; 310 Microwave radiation area; 400 Microwave source. Detailed Implementation

[0040] To better understand the above-mentioned objectives, features, and advantages of the present invention, the solutions of the present invention will be further described below. It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other.

[0041] Many specific details are set forth in the following description in order to provide a full understanding of the invention, but the invention may also be practiced in other ways different from those described herein; obviously, the embodiments in the specification are only some embodiments of the invention, and not all embodiments.

[0042] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0043] Example 1

[0044] Reference Figures 1 to 6 This embodiment provides a dual-mode neuromorphic magnetic navigation device, including a magnetic sensing substrate 100, a magnetic sensing pixel module 200, an array microwave antenna 300, and a microwave source 400.

[0045] The magnetic sensing substrate 100 incorporates a neural network architecture circuit and a signal conditioning circuit.

[0046] Specifically, the type of magnetic sensing substrate 100 is not limited. For example, in this embodiment, the magnetic sensing substrate 100 uses a customized PCB board, and the neural network architecture circuit and signal conditioning circuit are both integrated in this customized PCB board. The functions of the neural network architecture circuit and the signal conditioning circuit will be described in detail later.

[0047] The magnetic sensing pixel module 200 is provided in multiple arrays on the magnetic sensing substrate 100. Each magnetic sensing pixel module 200 includes a control substrate 210 and at least three fluorescence sensing units 220 disposed on the control substrate 210. The control substrate 210 has a built-in control circuit. Each fluorescence sensing unit 220 includes an NV color center 221, a filter 222 disposed on the light-emitting side of the NV color center 221, a photodetector 223 disposed on the light-emitting side of the filter 222, and a laser 224 arranged laterally toward the NV color center 221. The at least three fluorescence sensing units 220 are used to acquire signals from NV color centers 221 with different crystal orientations. The laser 224 and the photodetector 223 are both fixed on the control substrate 210 and electrically connected to the control circuit. The control substrate 210 is fixed on the magnetic sensing substrate 100, and the control circuit is electrically connected to the neural network architecture circuit and the signal conditioning circuit.

[0048] Specifically, the number of magnetic sensing pixel modules 200 is not limited; a larger number results in a more detailed magnetic image and higher detection accuracy, but also increases the computational load. In this embodiment, twenty-five magnetic sensing pixel modules 200 are arranged in a 5×5 square array on the magnetic sensing substrate 100, resulting in a more uniform magnetic image. In other embodiments, thirty magnetic sensing pixel modules 200 can also be designed, arranged in a 5×6 rectangular array on the magnetic sensing substrate 100.

[0049] Specifically, in this embodiment, the control substrate 210 is detachably plugged into the magnetic sensing substrate 100 via the electrical interface 230, which facilitates easy disassembly and replacement / maintenance. The electrical interface 230 can be a pin header type or other commonly used structures.

[0050] Specifically, the fixing methods of NV color center 221, filter 222, photodetector 223, and laser 224 are not limited. For example, in this embodiment, filter 222 is bonded to the underside of NV color center 221 by UV-curable optical adhesive, photodetector 223 is bonded to the underside of filter 222 by UV-curable optical adhesive, photodetector 223 is welded and fixed to control substrate 210, and laser 224 is fixed to control substrate 210 by aluminum nitride heat dissipation substrate and copper encapsulation structure located below it, so as to ensure that the operating temperature of laser 224 is within a safe range.

[0051] Specifically, in this embodiment, the laser 224 is a 532nm semiconductor laser, and the filter 222 is a 600nm-800nm ​​bandpass filter used to filter out the 532nm excitation light background. Since the NV color center 221 requires optical excitation with a 532nm laser, this embodiment uses a 532nm semiconductor laser as the laser 224. Because the NV color center 221 generates a fluorescence signal with a wavelength of 637nm after excitation, the filter 222 is a 600nm-800nm ​​bandpass filter, which can transmit the fluorescence band of the NV color center 221 while suppressing the excitation light band, and simultaneously reduces ambient light and other non-target wavelength light signals entering the photodetector 223.

[0052] Specifically, in this embodiment, the control substrate 210 is provided with four fluorescent sensing units 220 arranged in a cross shape, and the magnetic sensing substrate 100 is provided with a total of one hundred fluorescent sensing units 220.

[0053] The array microwave antenna 300 is arranged in parallel above the magnetic sensing substrate 100 and fixed relative to the magnetic sensing substrate 100. The array microwave antenna 300 is provided with multiple microwave radiation areas 310, which are arranged one-to-one with the NV color center 221 and are used to provide a microwave field for the NV color center 221.

[0054] Specifically, the method of fixing the array microwave antenna 300 relative to the magnetic sensing substrate 100 is not limited. For example, the array microwave antenna 300 can be directly fixed to the magnetic sensing substrate 100, or it can be fixed to an intermediate connector with the intermediate connector fixed relative to the magnetic sensing substrate 100.

[0055] Specifically, the number of microwave radiation regions 310 is not limited, but it must be equal to the number of NV color centers. In this embodiment, the array microwave antenna 300 is provided with one hundred microwave radiation regions 310.

[0056] Specifically, in this embodiment, the microwave radiation region 310 is a radiation window opened on the array microwave antenna 300, and the radiation window corresponds one-to-one with the position of the NV color center 221. The radiation window has two advantages: first, it is more conducive to the penetration of the microwave field; second, it is more conducive to the alignment and installation of the array microwave antenna 300 and the magnetic sensing substrate 100.

[0057] Specifically, in this embodiment, each microwave radiation region 310 of the array microwave antenna 300 is provided with an independent feeding network, and each feeding network sets its impedance parameters according to the responsivity weight of the corresponding fluorescence sensing unit 220. By regulating the microwave characteristics of the corresponding microwave radiation region 310 through the feeding network, the electron spin resonance process of the NV color center 221 can be controlled, thereby utilizing the reconfigurability of the NV color center 221 to adjust the polarity and magnitude of the responsivity of the fluorescence sensing unit 220.

[0058] More specifically, the power supply network adjusts the responsivity weight of the corresponding fluorescence sensing unit 220 by regulating the microwave frequency and / or power of each microwave radiation zone 310.

[0059] Among them, the microwave source 400 is connected to the array microwave antenna 300.

[0060] Specifically, the specifications of the microwave source 400 are not limited. For example, in this embodiment, the microwave source 400 is a small portable microwave source with a frequency range of 2.87 GHz ± 50 MHz, and is fixedly disposed relative to the magnetic sensing substrate 100 to be integrated into the magnetic navigation device.

[0061] Specifically, the number of microwave sources 400 is not limited. For example, in this embodiment, there are two microwave sources 400, both of which are connected to the array microwave antenna 300 via low-loss radio frequency connection lines.

[0062] The neural network architecture circuit is configured to directly perform a weighted summation operation on the output signal of the fluorescence sensing unit 220 in the analog domain and output navigation commands; the signal conditioning circuit is configured to selectively activate some of the fluorescence sensing units 220 to achieve magnetic compass mode or activate all of the fluorescence sensing units 220 to achieve magnetic map mode.

[0063] Specifically, the neural network architecture circuit of this embodiment adopts an integrated sensing and computing architecture, and constructs a hardware-based neural network forward propagation path based on Kirchhoff's current law. This circuit is used to multiply the output current of each fluorescence sensing unit 220 with the corresponding weight and then perform a summation operation, thereby directly completing matrix multiplication and addition operations in the analog domain, thus avoiding data transmission delay and power consumption loss caused by the separation of sensing and computing.

[0064] Specifically, the signal conditioning circuit in this embodiment includes a mode selection switch array, a signal amplification unit, and a signal output interface. The mode selection switch array is connected one-to-one with one hundred fluorescence sensing units 220, and is used to independently select or close the signal path of each fluorescence sensing unit 220 according to the working mode requirements. The signal amplification unit is used to pre-amplify the weak electrical signal after the fluorescence signal is converted, so as to improve the signal-to-noise ratio of signal detection. The signal output interface is used to transmit the output signal processed by the signal conditioning circuit to the neural network architecture circuit for further calculation and processing. In magnetic compass mode, the mode selection switch array only selects the three fluorescence sensing units 220 with orthogonal direction selectivity, and closes the signal path of the remaining ninety-seven fluorescence sensing units 220, thereby reducing system power consumption. In magnetic map mode, the mode selection switch array selects all one hundred fluorescence sensing units 220, so that the output signal of all sub-pixels is connected to the neural network architecture circuit for weighted summation.

[0065] The working principle of the dual-mode neuromorphic magnetic navigation device in this embodiment is as follows:

[0066] Laser 224 outputs 532nm excitation light and incident it onto NV color center 221 via lateral coupling. Simultaneously, microwave source 400 inputs microwave signal to array microwave antenna 300 through low-loss RF connection line. Each microwave radiation region 310 of array microwave antenna 300 provides microwave field for the corresponding NV color center 221. Under the combined action of laser excitation and microwave excitation, NV color center 221 generates fluorescence signal related to the external magnetic field strength. After the 532nm excitation light background is filtered out by 600nm-800nm ​​bandpass filter 222, the fluorescence signal is converted into a corresponding electrical signal output by photodetector 223.

[0067] The electrical signal output by the photodetector 223 is transmitted to the control circuit in the control substrate 210, and then to the signal conditioning circuit in the magnetic sensing substrate 100 through the electrical interface 230. The signal conditioning circuit independently selects or closes the signal path of each fluorescence sensing unit 220 through the mode selection switch array according to the requirements of the current working mode. In the magnetic compass mode, only three fluorescence sensing units 220 with different directional selectivity are activated, and the remaining ninety-seven fluorescence sensing units 220 remain in standby state to reduce system power consumption. In the magnetic map mode, all one hundred fluorescence sensing units 220 are activated to obtain complete magnetic map information.

[0068] The signal processed by the signal conditioning circuit is transmitted to the neural network architecture circuit. The neural network architecture circuit adopts an integrated sensing and computing architecture, which directly performs weighted summation on the output signals of each fluorescence sensing unit 220 in the analog domain. The responsivity weights of each fluorescence sensing unit 220 are configured through the feed network of the array microwave antenna 300 corresponding to the microwave radiation region 310. By adjusting the microwave frequency and / or power of each microwave radiation region 310, the responsivity weights of the corresponding fluorescence sensing unit 220 can be continuously adjusted, thereby completing the hardware mapping of the neural network weight matrix. Based on the above weighted summation results, the neural network architecture circuit outputs corresponding navigation commands to realize dual-mode cooperative navigation.

[0069] Example 2

[0070] This invention provides a dual-mode neuromorphic magnetic navigation method, which is implemented based on the dual-mode neuromorphic magnetic navigation device of Embodiment 1: When the distance between the real-time position and the target position is greater than a preset distance, navigation is performed using a magnetic compass mode; in magnetic compass mode, the signal conditioning circuit selects three fluorescent sensing units 220 with different directional selectivity to work through the control circuit, corresponding to the three-dimensional orthogonal coordinate axes respectively, to obtain XYZ three-axis magnetic vector information; when the distance between the real-time position and the target position is less than or equal to the preset distance, navigation is performed using a magnetic map mode; in magnetic map mode, the signal conditioning circuit controls all fluorescent sensing units 220 to work through the control circuit to obtain magnetic map information, and the neural network architecture circuit directly performs a weighted summation operation on the output signals of all fluorescent sensing units 220 in the analog domain, and outputs navigation commands based on the weighted summation result.

[0071] Specifically, in magnetic compass mode, the signal conditioning circuit controls three fluorescent sensing units 220 with different directional selectivity to be in working state through the control circuit, while the remaining fluorescent sensing units 220 are in standby state. The three fluorescent sensing units 220 with different directional selectivity sense the magnetic field components of the three axes of the three-dimensional orthogonal coordinate system respectively. By adjusting the microwave power of each fluorescent sensing unit 220 to make their responsivity consistent, the three fluorescent sensing units 220 respectively output electrical signals that are linearly related to the magnetic field strength of the corresponding axis. The three-axis magnetic field components are calculated based on the output signals of the three fluorescent sensing units 220, the vector information solution of the XYZ axes is established, the magnetic field direction navigation information is obtained, and the long-distance magnetic field direction positioning is completed.

[0072] Specifically, in magnetograph mode, the signal conditioning circuit controls all fluorescence sensing units 220 to be in working state through the control circuit. Each fluorescence sensing unit 220 senses the magnetic field strength information at its corresponding position and outputs the corresponding electrical signal. The electrical signals output by all fluorescence sensing units 220 are transmitted to the neural network architecture circuit for weighted summation. The weighted summation is directly completed in the analog domain based on Kirchhoff's current law. The responsivity weight of each fluorescence sensing unit 220 is configured through the feed network of the array microwave antenna 300 corresponding to the microwave radiation area 310. By adjusting the microwave frequency and / or power of each microwave radiation area 310, the responsivity weight of the corresponding fluorescence sensing unit 220 can be adjusted. Based on the weighted summation result, the current encoding signal corresponding to different magnetographs is output to realize magnetograph recognition based on the neuromorphic computing architecture, corresponding to various precise navigation commands, and completing short-range precise path navigation.

[0073] It should be noted that the dual-mode neuromorphic magnetic navigation method in this embodiment achieves a transition from long-distance coarse navigation to short-distance precise navigation by switching between magnetic compass mode and magnetic map mode. When the distance between the real-time position and the target position is greater than a preset distance, the magnetic compass mode is used to quickly determine the direction and perform approximate navigation. When the distance between the real-time position and the target position is less than or equal to the preset distance, the method switches to magnetic map mode for precise position identification and path planning. Through the above dual-mode cooperative navigation process, a continuous transition from coarse navigation to precise navigation is achieved, improving the overall navigation accuracy and robustness.

[0074] The effectiveness of this method will be verified through two experiments below.

[0075] Experiment 1: A target magnetic field is established below the device using a triaxial Helmholtz coil. Different magnitudes of target magnetic fields are created by varying the excitation current of the coil. The signal conditioning circuit selects three fluorescent sensing units 220 with different directional selectivity to operate. These three units correspond to the X, Y, and Z axes in a three-dimensional orthogonal coordinate system, respectively, to acquire the magnetic field components of the target magnetic field in these three orthogonal directions. The three fluorescent sensing units 220 detect the magnetic field components of the target magnetic field in the X, Y, and Z axes and output corresponding current values. These output current values ​​are then transmitted to an external signal processing module. The signal processing module calculates the vector magnetic field direction and compares it with the target magnetic field direction to verify the error between the orthogonal coordinate system magnetic field projection solution and the ideal parameter of the target magnetic field. This verifies the magnetic compass mode's ability to detect vector magnetic fields.

[0076] Experimental results: From Figure 7 It can be seen that the direction selectivity of the fluorescence sensing unit 220 enables rapid calculation of the vector magnetic field projected onto the orthogonal coordinate system. First, the microwave power of the three fluorescence sensing units 220 is adjusted to make their responsivity consistent (0.4 mA / Gs). The three fluorescence sensing units 220 selectively sense the axial magnetic information of the three-dimensional orthogonal coordinate system, and the angle between the target magnetic field direction and the three-dimensional orthogonal coordinate system is analyzed and calculated. Let be the angle between the direction of the vector magnetic field and the XY plane. Let be the angle between the vector magnetic field and the YZ plane. The projected components of the triaxial magnetic field pass through... This is used to calculate and ultimately combine weighting coefficients to represent the triaxial magnetic field strength using current information, thus enabling rapid acquisition of the magnetic field direction. For example... Figure 8As shown, by using a triaxial Helmholtz coil to independently input excitation currents to the X-axis, Y-axis and Z-axis respectively, the magnetic field components measured by each fluorescent sensing unit 220 are linearly characterized, demonstrating the linear detection capability of the three fluorescent sensing units 220 for the three-axis magnetic field components. At the same time, by comparing the measured vector magnetic field direction with the target magnetic field direction, it can be found that the vector magnetic field error is 0.05% to 6.35%, proving that the device has long-distance magnetic vector navigation capability.

[0077] Experiment 2: Constructing a real-world magnetic field navigation experimental platform. Four coil array targets with different magnetic field patterns were arranged within the experimental area to form magnetic field distribution patterns with spatially distinct characteristics, which were then used as the magnetic map information corresponding to different navigation targets. The magnetic map information corresponding to the four navigation targets was pre-trained offline using a computer-aided neural network algorithm to obtain the corresponding responsivity weight matrix. The magnetic sensor array consisted of twenty-five magnetic sensing pixel modules 200, each containing four fluorescent sensing units 220. By adjusting the microwave frequency and power of the array microwave antenna 300, the responsivity weight matrix was mapped to the responsivity and polarity of each fluorescent sensing unit 220. During the testing phase, the device was placed above the positions corresponding to the four different magnetic field patterns. Real-time magnetic field information was acquired through each magnetic sensing pixel module 200, and a weighted summation operation was performed based on the neural network hardware circuit to output the corresponding navigation commands. Finally, the magnetic map navigation capability of the device was verified by statistically analyzing the navigation accuracy of the recognition results of different magnetic field patterns under different noise conditions.

[0078] Experimental results: From Figure 9 It can be seen that there are twenty-five magnetic sensing pixel modules 200, and each magnetic sensing pixel module 200 consists of four fluorescence sensing units 220. The magnetic field strength sensed by each magnetic sensing pixel module 200 These are the numbers (1-25) for the magnetic sensing pixel modules 200. The weight distribution for each fluorescence sensing unit 220, The number assigned to each fluorescence sensing unit 220 is also the device output channel number. This device has four current output interfaces. For the bias resistor on each output path, Kirchhoff current is finally calculated using a neural network architecture. This completes the hardware mapping of the offline neural network algorithm, enabling the recognition and classification of four different magnetic field patterns. From Figure 10It can be seen that by performing labeled offline training on magnetic field patterns containing four typical different pixel distributions, and establishing the mapping relationship between magnetic field distribution and output current value, we characterize the output current values ​​of the four channels when the device identifies four different magnetic field patterns. It can be seen that when identifying the first "X" pattern, The current value is encoded as "1000", while the other three patterns are "0100", "0010", and "0001" respectively. The data shows that the current value has strong discriminative power and demonstrates high accuracy in recognizing the four magnetic field patterns. Finally, accuracy tests were conducted under different noise levels. It can be seen that at noise levels of 0.1-0.5, the device achieves a recognition accuracy close to 100%, verifying that the device still possesses magnetic map navigation capabilities in complex environments.

[0079] The above are merely specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Although detailed descriptions have been provided with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments, and they should all be covered within the protection scope of the claims.

Claims

1. A dual-mode neuromorphic magnetic navigation device, characterized in that, include: A magnetic sensing substrate (100) has a built-in neural network architecture circuit and a signal conditioning circuit; A magnetic sensing pixel module (200) is provided, which is arranged in an array on the magnetic sensing substrate (100). Each magnetic sensing pixel module (200) includes a control substrate (210) and at least three fluorescence sensing units (220) disposed on the control substrate (210). The control substrate (210) has a built-in control circuit. Each fluorescence sensing unit (220) includes an NV color center (221), a filter (222) disposed on the light-emitting side of the NV color center (221), and a light-emitting filter (222) disposed on the light-emitting side of the filter (222). The photodetector (223) and the laser (224) arranged laterally toward the NV color center (221), the at least three fluorescence sensing units (220) are used to acquire signals of NV color centers (221) with different crystal orientations respectively, the laser (224) and the photodetector (223) are both fixed on the control substrate (210) and are both electrically connected to the control circuit, the control substrate (210) is fixed on the magnetic sensing substrate (100), and the control circuit is electrically connected to the neural network architecture circuit and the signal conditioning circuit; An array microwave antenna (300) is arranged in parallel above the magnetic sensing substrate (100) and fixed relative to the magnetic sensing substrate (100). The array microwave antenna (300) is provided with a plurality of microwave radiation regions (310). The microwave radiation regions (310) are arranged one-to-one with the NV color center (221) and are used to provide a microwave field for the NV color center (221). A microwave source (400) is connected to the array microwave antenna (300); The neural network architecture circuit is configured to directly perform a weighted summation operation on the output signal of the fluorescence sensing unit (220) in the analog domain and output navigation commands; the signal conditioning circuit is configured to selectively activate some fluorescence sensing units (220) to achieve magnetic compass mode or activate all fluorescence sensing units (220) to achieve magnetic map mode.

2. The dual-mode neuromorphic magnetic navigation device according to claim 1, characterized in that, The control substrate (210) is detachably plugged into the magnetic sensing substrate (100) via an electrical interface (230).

3. The dual-mode neuromorphic magnetic navigation device according to claim 1, characterized in that, The control substrate (210) has four fluorescence sensing units (220) arranged in a cross shape.

4. The dual-mode neuromorphic magnetic navigation device according to claim 1, characterized in that, Multiple magnetic sensing pixel modules (200) are arranged in a square array on the magnetic sensing substrate (100).

5. The dual-mode neuromorphic magnetic navigation device according to claim 1, characterized in that, The laser (224) is a 532nm semiconductor laser, and the filter (222) is a 600nm-800nm ​​bandpass filter.

6. The dual-mode neuromorphic magnetic navigation device according to claim 1, characterized in that, Each microwave radiation region (310) of the array microwave antenna (300) is provided with an independent feeding network, and each feeding network sets impedance parameters according to the responsivity weight of the corresponding fluorescent sensing unit (220).

7. The dual-mode neuromorphic magnetic navigation device according to claim 6, characterized in that, The power supply network adjusts the responsivity weight of the corresponding fluorescence sensing unit (220) by regulating the microwave frequency and / or power of each microwave radiation region (310).

8. The dual-mode neuromorphic magnetic navigation device according to claim 1, characterized in that, The microwave radiation region (310) is a radiation window opened on the array microwave antenna (300), and the radiation window corresponds one-to-one with the position of the NV color center (221).

9. The dual-mode neuromorphic magnetic navigation device according to claim 1, characterized in that, The microwave source (400) is fixed relative to the magnetic sensing substrate (100), and the microwave source (400) is connected to the array microwave antenna (300) through a low-loss radio frequency connection line.

10. A dual-mode neuromorphic magnetic navigation method, characterized in that, It is based on the dual-mode neuromorphic magnetic navigation device according to any one of claims 1 to 9: When the distance between the real-time position and the target position is greater than the preset distance, the magnetic compass mode is used for navigation. In the magnetic compass mode, the signal conditioning circuit selects three fluorescent sensing units (220) with different directional selectivity to work through the control circuit, which correspond to the three-dimensional orthogonal coordinate axes respectively, in order to obtain XYZ three-axis magnetic vector information. When the distance between the real-time position and the target position is less than or equal to the preset distance, the magnetic map mode is used for navigation. In the magnetic map mode, the signal conditioning circuit controls all fluorescence sensing units (220) to work through the control circuit to obtain magnetic map information. The neural network architecture circuit directly performs weighted summation on the output signals of all fluorescence sensing units (220) in the analog domain and outputs navigation instructions based on the weighted summation result.

Images