A large-scale space brain navigation method based on high-dimensional cognitive coding

Abstract

The application discloses a large-scale space brain navigation method based on high-dimensional cognitive coding, which comprises the following steps: constructing grid cells by using an oscillation interference network with negative feedback, performing path integration on the three-axis speed of self-motion information, and realizing high-frequency rough estimation of the position of an unmanned system; constructing position cells by using a continuous attractor neural network, efficiently representing a large-scale space through a cyclic mapping mechanism, and realizing adaptive fusion of self-motion information and environmental information; and realizing memory and reasoning functions of the unmanned system on environmental information through an interest iteration mechanism for fast decoding of a cognitive map. The application combines self-motion information provided by an inertial navigation system and external environmental information provided by a global navigation satellite system and a camera, and provides an efficient and reliable large-scale space brain navigation method for the unmanned system.

Description

Technical Field

[0001] This invention belongs to the field of intelligent navigation technology, specifically relating to a large-scale spatial brain-like navigation method based on high-dimensional cognitive coding. Background Technology

[0002] The application of unmanned systems is rapidly expanding into various fields, becoming a significant driving force for modern technological development. With continuous technological advancements, unmanned systems are widely used in environmental monitoring, military reconnaissance, search and rescue, and other scenarios. These applications typically require high-precision and high-reliability navigation and positioning capabilities. Especially in large-scale spaces, unmanned systems face complex and ever-changing environments and geographical features, making the reliability and accuracy of their navigation technology a core challenge.

[0003] Existing multi-source information fusion methods mainly include filtering and optimization techniques. Filtering-based methods rely on the covariance matrix for state estimation, and their assumptions of first-order Markovianness and Gaussian noise lead to performance degradation in complex environments. Optimization-based methods are prone to getting trapped in local optima and have high computational complexity, making them unsuitable for real-time applications in complex, large-scale scenarios. Furthermore, the accuracy of error modeling decreases when sensors are disturbed, directly affecting the precision of the fusion results.

[0004] In contrast, mammals such as bats exhibit stable and reliable navigation capabilities in large-scale, complex environments. Their navigation does not rely on precise external modeling, but rather on high-dimensional semantic cognition of the external environment and adaptive fusion of multimodal information. In particular, grid cells and place cells in the hippocampus-entorhinal cortex circuit play a central role in spatial navigation and memory. Through the synergistic action of these cells, a cognitive map can be constructed through self-motor perception and environmental cognition, enabling reliable and stable navigation.

[0005] Therefore, researching brain-like navigation methods based on high-dimensional cognitive coding to improve the robustness, intelligence, and autonomy of unmanned system navigation in large-scale complex spaces has important scientific and applied value. Summary of the Invention

[0006] Purpose of the invention: This invention proposes a large-scale spatial brain-like navigation method based on high-dimensional cognitive coding, which solves the problems of existing navigation methods relying on precise mathematical models in large-scale spaces and failing to fully utilize high-dimensional cognitive information in the external perception environment, thereby improving the reliability, stability and intelligence of navigation systems in unmanned systems in large-scale complex environments.

[0007] Technical Solution: The large-scale spatial brain-like navigation method based on high-dimensional cognitive coding described in this invention specifically includes the following steps:

[0008] (1) Acquire self-motion information and external perception information based on unmanned systems;

[0009] (2) Construct a cognitive map model, a grid cell model, and a place cell model; the cognitive map is used to provide an internal representation of external cognitive information; the grid cell is used to simulate the path integral function in the brain; the place cell is used to fuse self-motor information and external perception information;

[0010] (3) Extract semantic pointers from images in external perception information and perform similarity measurement with cognitive maps; if a similar scene is found, proceed to step (4); otherwise, proceed to step (5);

[0011] (4) Decode the spatial semantic pointers of similar scenes in step (3) based on interest-based iterative cognitive map, output the localization results of the unmanned system, update the initial value of the grid cell path integral, and execute step (8).

[0012] (5) The three-axis velocity in the self-motion information is encoded by the grid cell, the path integral is realized by the activity iteration update, and the high-frequency position estimate of the unmanned system is obtained by decoding the discharge result of the grid cell model.

[0013] (6) Encode the estimated information decoded in step (5) and the location information provided in the external perception information using the location cells, update the activity and decode the output location information after the unmanned system integrates multi-source information, bind it with the spatial semantic pointer extracted in step (3) and memorize it in the cognitive map;

[0014] (7) If the wave packet moves to the edge of the position cell representation range, the position cell is remapped using the cyclic mapping mechanism;

[0015] (8) Return to step (3) and loop until the navigation ends.

[0016] Furthermore, the self-motion information in step (1) includes three-axis velocity information provided by an inertial sensor; the external perception information includes images provided by a camera, visual odometry information, and position information provided by a global navigation satellite system.

[0017] Furthermore, the process of constructing the cognitive map model in step (2) is as follows:

[0018] To achieve high-dimensional representation of continuous spatial information, the mapping relationship from Euclidean spatial location vectors to cognitive maps is established. This mapping is encoded in the following way:

[0019] Ω(x)=f -1 {e iAx}

[0020] Where x is a position vector in Euclidean space. Let f be the encoding matrix. -1 For the inverse Fourier transform, to ensure that the vector Ω in the cognitive map is real, we must choose a value such that e iAx Matrices with conjugate symmetry:

[0021]

[0022] in, This indicates a bundling operation. It represents the Hadamard product; through one or more binding operations, it enables the continuous movement of an unmanned system in space.

[0023] Furthermore, the process of implementing the network-based lattice cell model described in step (2) is as follows:

[0024] Velocity information provided by motion information in Euclidean space Transformed into the time derivative of spatial semantic pointers in a cognitive map

[0025]

[0026] The transformation matrix A is composed of multiple sets of row vectors, and each set of vectors k must satisfy the following:

[0027]

[0028] Through the mutual oscillation and interference of multiple sets of vectors with a phase difference of 120°, the final grid cell firing pattern will present a grid-like shape in Euclidean space.

[0029] Furthermore, the process of constructing the cell model described in step (2) is as follows: membrane potential u PC The update process for (m,t) is as follows:

[0030]

[0031] Where, τ PC r is the time constant of the synapse at the location of the cell. PC (n,t) represents the firing rate of the position cell. The magnitude of the current encoded by the position of the unmanned system, γ PC J is the global suppression constant. PC (m,n) represents the connection weights between positional cells, ρ PC denoted as the representation density of the location cell network, where m is the cell number at the current location and n is the cell number at other locations;

[0032] The connection weights between location cells are determined by the standard deviation α. PC The Gaussian function determines:

[0033]

[0034] The firing rate r of the position cell PC The calculation method for (n,t) is as follows:

[0035]

[0036] Where, k PC This is the adjustment coefficient.

[0037] Furthermore, the implementation process of step (3) is as follows:

[0038] To extract semantic pointers from images, a convolutional neural network is used to extract features from the current frame image, and the network output, which is a b-dimensional vector, is normalized to obtain the feature semantic pointers. Meanwhile, other semantic information perceived in this frame is represented using the orthogonal vectors of a b-dimensional hypersphere:

[0039]

[0040] in For the Euclidean norm, δ ij If the keyframe is a Kronecker symbol, then the joint representation of its semantic pointers is calculated as follows:

[0041]

[0042] During the comparison with the cognitive map M, the external perception information of this frame will be inverted to obtain the comparison result:

[0043]

[0044] The results are judged. If ||p||2 < ε, it can be considered that there is no relevant memory in the currently stored cognitive map. Otherwise, it is considered that a similar scene has been found and the spatial semantic pointer p0 in the scene has been obtained. Further decoding is needed to obtain the position of the unmanned system in Euclidean space.

[0045] Furthermore, the implementation process of step (4) is as follows:

[0046] For a large-scale space S0 of length c and width d, decoding the semantic pointer p0 of the target space requires dividing it into r×s regions of interest, satisfying the following requirements:

[0047]

[0048] Where r, s ∈ N, and t is a positive constant; then, each interest point is encoded to obtain its corresponding semantic feature vector. Next, cosine similarity is calculated with the target space semantic pointer p0 to identify the cognitive space region most similar to the target location:

[0049]

[0050] At this point, the decoding result is the position with the highest cosine similarity to the semantic pointer p0 in the target space, which is the output of the first decoding. Iterative updates continue within the identified region of interest S1, and the result of the w-th iteration is used to calculate:

[0051]

[0052] Where C is the inference threshold, if the above formula is satisfied, the location information of the scene can be inferred.

[0053] Furthermore, the implementation process of step (5) is as follows:

[0054] The update and iteration process for changes in the activity of grid cells is as follows:

[0055]

[0056] Where j is the Fourier component, r j =|f{Ω(x)} j | is the feedback factor. The frequency is time-varying on the unit circle; the model realizes grid-like discharge within a neural engineering framework and achieves path integration through oscillatory interference of different phase information.

[0057] Furthermore, the implementation process of step (6) is as follows:

[0058] For a cell network representing a location range of [c0, d0], the location information obtained from the grid cell path integral and the location information from external perception will be encoded using radial basis functions:

[0059]

[0060] in, The total input after multi-sensor encoding. This represents the absolute difference between the position represented by the position cell and the position measurement information of the i-th sensor. Let i be the encoding weight of the i-th sensor. The standard deviation of the encoding for the i-th sensor;

[0061] After the network has been iterated and updated to a stable state, its location cell fusion information is decoded:

[0062]

[0063] in, Let P be the position represented by the i-th position cell, l be the spatial length represented by the position cell, and P be the position of the i-th position cell. PC (t) represents the location information of the unmanned system decoded at time t;

[0064] During the spatial semantic pointer binding process, the spatial semantic pointer p of the cognitive map M and the current scene are... sem The memory process can then be described as follows:

[0065]

[0066] The current spatial semantic pointer is combined with the location of the unmanned system through a binding operation and then memorized into the cognitive map.

[0067] Furthermore, the implementation process of step (7) is as follows:

[0068] During the location-based cell update process based on cyclic mapping, the change in location-based cell membrane potential is as follows:

[0069] u PC (m,t)=u PC ((m-h+N)modN,t)

[0070] Where h is the number of cells mapped and translated, and N is the total number of position cells; the spatial range represented by the position cells also changes as follows:

[0071]

[0072] Among them, c n and d n The spatial extent represented by the position cell after the nth decoding.

[0073] Beneficial Effects: Compared with existing technologies, the beneficial effects of this invention are as follows: Based on the navigation and cognitive mechanisms of mammals such as bats, this invention utilizes high-dimensional cognitive encoding to achieve efficient utilization of external perceptual information, possessing advantages such as high computational efficiency, strong autonomy, and high level of intelligence; it constructs a cognitive map using spatial semantic pointers, and can improve the robustness and reliability of the navigation system by using historical memory scenarios to assist in correcting the current motion estimation of the unmanned system; it improves the efficiency and accuracy of cognitive map decoding by using an interest-based iterative recall mechanism to rapidly update the cognitive map; it achieves large-scale spatial representation with a small number of neurons through cyclic mapping of position cells, while maintaining the same representation resolution; this invention does not require linearized perceptual information or Gaussian noise assumptions, and fully utilizes high-dimensional perceptual information in the external environment, possessing advantages such as strong universality of navigation architecture and stable and reliable navigation results, and can be used for intelligent autonomous navigation of unmanned systems in large-scale spaces. Attached Figure Description

[0074] Figure 1 This is a flowchart of the present invention;

[0075] Figure 2 This is a schematic diagram of cognitive map decoding based on an interest-based iterative mechanism;

[0076] Figure 3 This is a schematic diagram of position cell iterative update based on cyclic mapping;

[0077] Figure 4 This is a comparison chart of experimental results. Detailed Implementation

[0078] The present invention will now be described in further detail with reference to the accompanying drawings.

[0079] like Figure 1 As shown, this invention proposes a large-scale spatial brain-like navigation method based on high-dimensional cognitive encoding. This method involves high-dimensional cognitive encoding of external sensory information and intelligently fusing it with self-motion information through position cells. If the same scene is encountered again, the location is directly decoded from the cognitive map memory, obtaining a stable and reliable positioning result. Specifically, it includes the following steps:

[0080] Step 1: The sensors of the unmanned system include inertial sensors, cameras, and global navigation satellite systems. The self-motion information includes three-axis velocity information provided by the inertial sensors, and the external perception information includes images provided by the cameras, visual odometry information, and position information provided by the global navigation satellite systems.

[0081] Step 2: Construct a cognitive map model, a grid cell model, and a place cell model; the cognitive map is used to provide an internal representation of external cognitive information; the grid cells are used to simulate the path integral function in the brain; and the place cells are used to fuse self-motor information and external perception information.

[0082] In the process of constructing a cognitive map model, in order to achieve a high-dimensional representation of continuous spatial information, a mapping relationship is established from Euclidean space location vectors to the cognitive map. This mapping can be encoded in the following way:

[0083]

[0084] Where x is a position vector in Euclidean space. Let f be the encoding matrix. -1 For the inverse Fourier transform, to ensure that the vector Ω in the cognitive map is real, we must choose a value such that e iAx Matrices with conjugate symmetry:

[0085]

[0086] in, This indicates the defined bundling operation. This represents the Hadamard product, an operation that allows for direct updating of vectors in the cognitive map without repeated decoding. Through one or more binding operations, continuous movement of the unmanned system in space is achieved.

[0087] During the construction of the grid cell model, velocity information provided by motion information in Euclidean space is used. Transformed into the time derivative of spatial semantic pointers in a cognitive map

[0088]

[0089] The transformation matrix A is composed of multiple sets of row vectors, and each set of vectors k must satisfy the following:

[0090]

[0091] Through the mutual oscillation and interference of multiple sets of vectors with a phase difference of 120°, the final grid cell firing pattern will present a grid-like shape in Euclidean space.

[0092] During the construction of the location cell model, the membrane potential u PC The update process for (m,t) is as follows:

[0093]

[0094] Where, τ PC r is the time constant of the synapse at the location of the cell. PC (n,t) represents the firing rate of the position cell. The magnitude of the current encoded by the position of the unmanned system, γ PC J is the global suppression constant. PC (m,n) represents the connection weights between location cells, ρ PC denoted as the representation density of the location cell network, where m is the cell number at the current location and n is the cell number at other locations.

[0095] The connection weights between location cells are determined by the standard deviation α. PC The Gaussian function determines:

[0096]

[0097] The firing rate r of the position cell PC The calculation method for (n,t) is as follows:

[0098]

[0099] Where, k PC This is the adjustment coefficient.

[0100] Step 3: Extract semantic pointers from images in external perception information and perform similarity measurement with the cognitive map; if a similar scene is found, proceed to step 4; otherwise, proceed to step 5.

[0101] In the process of extracting image semantic pointers, a convolutional neural network is used to extract features of the current frame image, and the network output, which is a b-dimensional vector, is normalized to obtain the feature semantic pointers. At the same time, other semantic information perceived about the frame (such as color) object (etc.) Characterized using the orthogonal vectors of the b-dimensional hypersphere:

[0102]

[0103] in For the Euclidean norm, δ ij Let it be a Kronecker symbol. Then the semantic pointer of this keyframe can be jointly represented and calculated:

[0104]

[0105] During the comparison with the cognitive map M, the external perception information of this frame will be inverted to obtain the comparison result:

[0106]

[0107] The results are judged. If ||p||2 < ε (ε is a constant), it can be considered that there is no relevant memory in the currently stored cognitive map. Otherwise, it is considered that a similar scene has been found and the spatial semantic pointer p0 in the scene has been obtained. Further decoding is needed to obtain the position of the unmanned system in Euclidean space.

[0108] Step 4: Decode the spatial semantic pointers of similar scenes in Step 3 using interest-based iterative cognitive map decoding, output the localization results of the unmanned system, update the initial values ​​of the grid cell path integrals, and then execute Step 8.

[0109] In the process of decoding similar scenes in a cognitive map based on interest iteration, for a large-scale space S0 with length c and width d, decoding the semantic pointer p0 of the target space requires dividing it into r×s interest regions, which must satisfy the following:

[0110]

[0111] Where r, s ∈ N, and t is a positive constant. Then, each interest point is encoded to obtain its corresponding semantic feature vector. Next, cosine similarity is calculated between the target location p0 and the target location p0 to identify the cognitive space region most similar to the target location. This process is achieved using the following formula:

[0112]

[0113] At this point, the decoding result is the position with the highest cosine similarity to the target position p0, which is the output of the first decoding. During this process, the decoding error is controlled within t / 2. As the region of interest becomes increasingly precise, it is no longer necessary to compare the entire space one by one; instead, iterative updates can continue only within the identified region of interest S1. Taking the result of the w-th iteration, calculate:

[0114]

[0115] Where C is the inference threshold, if the above formula is satisfied, the location information of the scene can be inferred.

[0116] Step 5: Encode the three-axis velocity in the self-motion information using grid cells, and realize the path integral through active iterative update; obtain a relatively coarse high-frequency position estimate of the unmanned system by decoding the discharge results of the grid cell model.

[0117] The iterative process by which grid cells produce changes in activity is as follows:

[0118]

[0119] Where j is the Fourier component, r j =|f{Ω(x)} j | is the feedback factor. The frequency is time-varying on the unit circle. This model, implemented within a neural engineering framework, can exhibit grid-like discharges and achieve path integration through oscillatory interference of different phase information.

[0120] The grid cell decoding process is the same as the spatial semantic pointer decoding method in step 4 above.

[0121] Step 6: Encode the estimated information decoded in Step 5 and the location information provided by external perception information using the location cells, update the activity and decode the output location information of the unmanned system after fusing multi-source information, bind it with the spatial semantic pointer extracted in Step 3, and memorize it in the cognitive map.

[0122] For a cell network representing a location range of [c0, d0], the location information obtained from the grid cell path integral and the location information from external perception will be encoded using radial basis functions:

[0123]

[0124] in, The total input after multi-sensor encoding. This represents the absolute difference between the position represented by the position cell and the position measurement information of the i-th sensor. Let i be the encoding weight of the i-th sensor. Let be the standard deviation of the encoding for the i-th sensor.

[0125] After the network has been iterated and updated to a stable state, its location cell fusion information is decoded:

[0126]

[0127] in, Let P be the position represented by the i-th position cell, l be the spatial length represented by the position cell, and P be the position of the i-th position cell. PC (t) represents the location information of the unmanned system decoded at time t.

[0128] During the spatial semantic pointer binding process, the spatial semantic pointer p of the cognitive map M and the current scene are... sem The memory process can then be described as follows:

[0129]

[0130] The current spatial semantic pointer is combined with the location of the unmanned system through a binding operation and then memorized into the cognitive map.

[0131] Step 7: If the wave packet moves close to the edge of the position cell representation range, the position cell is remapped using the cyclic mapping mechanism.

[0132] The positional cell membrane potential change is as follows:

[0133] u PC (m,t)=u PC ((m-h+N)modN,t)

[0134] Where h represents the number of cells mapped and translated, and N represents the total number of position cells. Through this operation, the spatial extent represented by the position cells also changes as follows:

[0135]

[0136] Among them, c n and d n The spatial extent represented by the position cell after the nth decoding.

[0137] Step 8: Continue repeating the above steps from Step 3 until the navigation ends.

[0138] Figure 2 This diagram illustrates the cognitive map decoding process based on interest iteration, where a and b represent the size of the cognitive map, p0 is a pointer to the high-dimensional semantics to be decoded in the cognitive map, and S0 is the entire region of the cognitive map. For the nth interest region iteration, the (1,1)th region is used to achieve fast and accurate decoding of large-scale spatial cognitive maps through interest iteration. Figure 3 This is a schematic diagram of the position cell update process based on cyclic mapping. When a wave packet is detected to be near the edge of the representation range, the wave packet is mapped to the middle region of the representation range by remapping the position cell membrane potential. At the same time, the spatial representation range is updated, realizing efficient representation of large-scale space and accurate fusion of multi-sensor information. Figure 4 To compare the experimental results of different model algorithms, the present invention was used for repeated tests, as shown in the red part of the figure. It can be clearly seen from the figure that the present method has good stability and accuracy in large-scale spatial navigation and positioning.

[0139] The present invention has been described in detail above with reference to specific embodiments; however, these descriptions should not be construed as limiting the present invention. Those skilled in the art will understand that various equivalent substitutions, modifications, or improvements can be made to the technical solutions and embodiments of the present invention without departing from the spirit and scope of the invention, and all such modifications and improvements fall within the scope of the present invention. The scope of protection of the present invention is defined by the appended claims.

Claims

1. A large-scale spatial brain-like navigation method based on high-dimensional cognitive coding, characterized in that, Includes the following steps: (1) Acquire self-motion information and external perception information based on unmanned systems; (2) Construct a cognitive map model, a grid cell model, and a place cell model; the cognitive map is used to provide an internal representation of external cognitive information; the grid cell is used to simulate the path integral function in the brain; the place cell is used to fuse self-motor information and external perception information; (3) Extract semantic pointers from images in external perception information and perform similarity measurement with cognitive maps; if a similar scene is found, proceed to step (4); otherwise, proceed to step (5); (4) Decode the spatial semantic pointers of similar scenes in step (3) based on interest-based iterative cognitive map, output the localization results of the unmanned system, update the initial value of the grid cell path integral, and execute step (8). (5) The three-axis velocity in the self-motion information is encoded by the grid cell, the path integral is realized by the activity iteration update, and the high-frequency position estimate of the unmanned system is obtained by decoding the discharge result of the grid cell model. (6) Encode the estimated information decoded in step (5) and the location information provided in the external perception information using the location cells, update the activity and decode the output location information after the unmanned system integrates multi-source information, bind it with the spatial semantic pointer extracted in step (3) and memorize it in the cognitive map; (7) If the wave packet moves to the edge of the position cell representation range, the position cell is remapped using the cyclic mapping mechanism; (8) Return to step (3) and loop until the navigation ends.

2. The large-scale spatial brain-like navigation method based on high-dimensional cognitive coding according to claim 1, characterized in that, The self-motion information in step (1) includes three-axis velocity information provided by the inertial sensor; the external perception information includes images provided by the camera, visual odometry information, and position information provided by the global navigation satellite system.

3. The large-scale spatial brain-like navigation method based on high-dimensional cognitive coding according to claim 1, characterized in that, The process of constructing the cognitive map model in step (2) is as follows: To achieve high-dimensional representation of continuous spatial information, the mapping relationship from Euclidean spatial location vectors to cognitive maps is established. This mapping is encoded in the following way: Ω(x)=f -1 {e iAx } Where x is a position vector in Euclidean space. Let f be the encoding matrix. -1 For the inverse Fourier transform, to ensure that the vector Ω in the cognitive map is real, we must choose a value such that e iAx Matrices with conjugate symmetry: in, This indicates a bundling operation. It represents the Hadamard product; through one or more binding operations, it enables the continuous movement of an unmanned system in space.

4. The large-scale spatial brain-like navigation method based on high-dimensional cognitive coding according to claim 1, characterized in that, The process of constructing the grid cell model in step (2) is as follows: Velocity information provided by motion information in Euclidean space Transformed into the time derivative of spatial semantic pointers in a cognitive map The transformation matrix A is composed of multiple sets of row vectors, and each set of vectors k must satisfy the following: Through the mutual oscillation and interference of multiple sets of vectors with a phase difference of 120°, the final grid cell firing pattern will present a grid-like shape in Euclidean space.

5. The large-scale spatial brain-like navigation method based on high-dimensional cognitive coding according to claim 1, characterized in that, The process of constructing the cell model in step (2) is as follows: membrane potential u PC The update process for (m,t) is as follows: Where, τ PC r is the time constant of the synapse at the location of the cell. PC (n,t) represents the firing rate of the position cell. The magnitude of the current encoded by the position of the unmanned system, γ PC J is the global suppression constant. PC (m,n) represents the connection weights between location cells, ρ PC denoted as the representation density of the location cell network, where m is the cell number at the current location and n is the cell number at other locations; The connection weights between location cells are determined by the standard deviation α. PC The Gaussian function determines: The firing rate r of the position cell PC The calculation method for (n,t) is as follows: Where, k PC This is the adjustment coefficient.

6. The large-scale spatial brain-like navigation method based on high-dimensional cognitive coding according to claim 1, characterized in that, The implementation process of step (3) is as follows: To extract semantic pointers from images, a convolutional neural network is used to extract features from the current frame image, and the network output, which is a b-dimensional vector, is normalized to obtain the feature semantic pointers. Meanwhile, other semantic information perceived in this frame is represented using the orthogonal vectors of a b-dimensional hypersphere: in For the Euclidean norm, δ ij If the keyframe is a Kronecker symbol, then the joint representation of its semantic pointers is calculated as follows: During the comparison with the cognitive map M, the external perception information of this frame will be inverted to obtain the comparison result: The results are judged. If ||p||2 < ε, it is considered that there is no relevant memory in the currently stored cognitive map. Otherwise, it is considered that a similar scene has been found and the spatial semantic pointer p0 in the scene has been obtained. Further decoding is needed to obtain the position of the unmanned system in Euclidean space.

7. The large-scale spatial brain-like navigation method based on high-dimensional cognitive coding according to claim 1, characterized in that, The implementation process of step (4) is as follows: For a large-scale space S0 of length c and width d, decoding the semantic pointer p0 of the target space requires dividing it into r×s regions of interest, satisfying the following requirements: Where r, s ∈ N, and t is a positive constant; then, each interest point is encoded to obtain its corresponding semantic feature vector. Next, cosine similarity is calculated with the target space semantic pointer p0 to identify the cognitive space region most similar to the target location: At this point, the decoding result is the position with the highest cosine similarity to the semantic pointer p0 in the target space, which is the output of the first decoding. Iterative updates continue within the identified region of interest S1, and the result of the w-th iteration is used to calculate: Where C is the inference threshold, if the above formula is satisfied, the location information of the scene is obtained.

8. The large-scale spatial brain-like navigation method based on high-dimensional cognitive coding according to claim 1, characterized in that, The implementation process of step (5) is as follows: The update and iteration process for changes in the activity of grid cells is as follows: Where j is the Fourier component, r j =|f{Ω(x)} j | is the feedback factor. The frequency is time-varying on the unit circle; the model realizes grid-like discharge within a neural engineering framework and achieves path integration through oscillatory interference of different phase information.

9. A large-scale spatial brain-like navigation method based on high-dimensional cognitive coding according to claim 1, characterized in that, The implementation process of step (6) is as follows: For a cell network representing a location range of [c0, d0], the location information obtained from the grid cell path integral and the location information from external perception will be encoded using radial basis functions: in, The total input after multi-sensor encoding. This represents the absolute difference between the position represented by the position cell and the position measurement information of the i-th sensor. Let i be the encoding weight of the i-th sensor. The standard deviation of the encoding for the i-th sensor; After the network has been iterated and updated to a stable state, its location cell fusion information is decoded: in, Let P be the position represented by the i-th position cell, l be the spatial length represented by the position cell, and P be the position of the i-th position cell. PC (t) represents the location information of the unmanned system decoded at time t; During the spatial semantic pointer binding process, the spatial semantic pointer p of the cognitive map M and the current scene are... sem The memory process can be described as follows: The current spatial semantic pointer is combined with the unmanned system's location and memorized into the cognitive map through a binding operation.

10. A large-scale spatial brain-like navigation method based on high-dimensional cognitive coding according to claim 1, characterized in that, The implementation process of step (7) is as follows: During the location-based cell update process based on cyclic mapping, the change in location-based cell membrane potential is as follows: u PC (m,t)=u PC ((m-h+N)modN,t) Where h is the number of cells mapped and translated, and N is the total number of position cells; the spatial range represented by the position cells also changes as follows: Among them, c n and d n The spatial extent represented by the position cell after the nth decoding.

Images