Homing route AI prediction and big data optimization system and method for homing pigeon locator

Abstract

The invention discloses a homing route AI prediction and big data optimization system and method for a homing pigeon locator, and relates to the technical field of homing pigeon homing route intelligent prediction. An external environment data receiving module; the knowledge migration technology module is used for reconstructing the heterologous biological navigation data into neural excitation signals; the virtual geographic information system module is used for generating candidate homing path segments based on the neural excitation signals; the space-time convolution model module is used for extracting group flight decision features based on the historical trajectory of the homing pigeon group; the chaos convergence controller module is used for analyzing the stability of the path fragment; and the topological neural network module is used for synthesizing the stable fragments into a continuous homing navigation instruction set. According to the homing navigation system, the biosensor and the AI technology are integrated, physiological and environmental data of the homing pigeons are collected in real time, neural excitation signals are reconstructed, the path is optimized, homing navigation precision is remarkably improved, energy consumption is reduced, and turbulence and interference resistance is enhanced.

Description

Technical Field

[0001] This invention relates to the field of AI-enabled big data intelligent prediction technology for homing pigeon homing routes, and in particular to an AI prediction and big data optimization system and method for homing pigeon locators. Background Technology

[0002] Homing pigeons rely on their innate multiple navigation mechanisms for homing, including the ability to sense the strength of the Earth's magnetic field, the ability to identify the sun's position and polarized light, and the ability to visually remember complex surface features. However, traditional human experience and basic electronic tracking equipment have significant limitations in route prediction: First, sudden geomagnetic disturbances in the natural environment, spatial orientation confusion caused by dense buildings, and changes in flight drag caused by extreme weather can all seriously interfere with the pigeons' orientation judgment, causing fluctuations or even sharp drops in homing rates; Second, although existing positioning devices can record discrete trajectory points, they cannot dynamically integrate environmental variables and biological behavioral characteristics, and lack the ability to predict homing routes; Third, the process of acquiring training data is susceptible to contamination by human cheating, and the fragmented storage of heterogeneous data further hinders the construction of high-precision prediction models. Summary of the Invention

[0003] In view of this, the present invention proposes an AI prediction and big data optimization system and method for homing pigeon locators to solve the problem of difficulty in accurately predicting homing pigeon routes in the prior art.

[0004] The specific technical solution of this invention is as follows: An AI-based prediction and big data optimization system for homing routes in pigeon locators includes: A miniature biosensor module, installed inside the pigeon locator, is used to collect real-time physiological data of the pigeon's flight. External environment data receiving module, used to acquire meteorological and topographic data; The signal fusion module is used to integrate physiological and environmental data to form a dynamic sensing dataset; The knowledge transfer technology module is used to reconstruct heterogeneous biological navigation data into neural excitation signals; The virtual geographic information system module is used to generate candidate homing path fragments based on neural excitation signals; The spatiotemporal convolutional model module is used to extract group flight decision features based on the historical trajectories of homing pigeon flocks; Chaotic convergence controller module, used to analyze the stability of path segments; The topological neural network module is used to synthesize a continuous homing navigation instruction set from stable segments.

[0005] Specifically, the signal fusion module uses a time alignment algorithm to eliminate device latency differences and applies a moving average algorithm to align timing, ensuring that biosignals and environmental data are accurately matched within a millisecond-level time window, and outputting a structured dataset updated once per second; the data fields include timestamps, signal values ​​and spatial coordinates, and an automatic retransmission mechanism ensures data integrity, triggering retransmission when the signal loss rate exceeds a preset threshold.

[0006] Specifically, the knowledge transfer technology module includes a generative adversarial network to reconstruct neural excitation signals, with the generator being a three-layer fully connected neural network and the discriminator being a convolutional neural network; the reconstructed signal amplitude is normalized and candidate path segments are generated through a virtual geographic information system, with the length of each path segment including terrain avoidance strategies and energy consumption estimates; the migration process pre-verifies species physiological compatibility to ensure that the neuron activation frequency matching degree is higher than a threshold.

[0007] Specifically, the chaotic convergence controller module analyzes the stability of path segments in phase space based on the Lyapunov index and sets a biological stability threshold to screen stable segments; if the stability index is less than the preset value, it is considered to meet the requirements, and the screening process ensures that the homing pigeons' homing path reduces flight risks.

[0008] Specifically, the topological neural network module adopts a fully connected heuristic search rule to connect stable path segments end to end to generate a continuous homing navigation instruction set with a navigation point every 100 meters; at the same time, it avoids electromagnetic interference sources and predator activity hotspots, and the output instruction set conforms to the principle of shortest time optimization, ensuring that the homing pigeons return to their nests safely and efficiently.

[0009] A method for AI prediction and big data optimization of homing routes for pigeon locators includes: By integrating a micro biosensor with an external environment data receiving module, the system collects real-time data on pigeon physiological tremor waveforms, geomagnetic induction intensity, visual spectral characteristics, and atmospheric turbulence terrain contours. A time alignment algorithm is then used to fuse these data to form a dynamic sensing dataset. Through knowledge transfer technology, the activation patterns of celestial navigation neurons of migratory birds and the geomagnetic memory maps of marine fish are reconstructed into resolvable neural excitation signals for homing pigeons, and candidate homing path fragments including terrain avoidance strategies and energy consumption estimates are generated in a virtual geographic information system. A spatiotemporal convolutional model was established using the historical homing trajectories of homing pigeons to extract the group's flight decision-making features. Combined with a micro barometer to perceive the direction of micro airflow vortices in real time, the path segments were dynamically corrected to generate anti-turbulence interference flight vector sequences. By analyzing the stability of fragments based on the Lyapunov exponent using a chaotic convergence controller and selecting fragments that meet the biological homeostasis threshold, a topological neural network is used to connect the fragments end to end to synthesize a continuous homing navigation instruction set, while avoiding electromagnetic interference sources and enemy activity hotspots.

[0010] Specifically, real-time collection of homing pigeon physiological data includes using a miniature triaxial accelerometer to detect physiological tremor waveforms, a magnetometer to measure geomagnetic induction intensity, a spectrometer to capture visual spectral features, and a meteorological satellite data receiving module to analyze atmospheric turbulence intensity and terrain contours; the data acquisition frequency is no less than one hundred times per second, and real-time recording is performed when the homing pigeon accelerates or turns to avoid signal loss.

[0011] Specifically, the reconstructed neural excitation signal is achieved through a generative adversarial network. The generator is input with migratory bird neuron patterns, and the discriminator verifies that the signal conforms to the biological constraints of homing pigeons. The reconstructed signal is used to drive the path generation algorithm in the virtual geographic information system, automatically avoid steep slope obstacles with a slope greater than a preset angle, and calculate the estimated energy consumption based on the bird flight dynamics formula.

[0012] Specifically, the dynamic path correction includes: establishing a spatiotemporal convolution model to extract population density distribution characteristics, sensing the vortex direction through a micro barometer, and using the viscous fluid resistance equation to calculate the heading correction and generate an anti-turbulence interference flight vector sequence.

[0013] Specifically, to avoid electromagnetic interference sources and predator activity hotspots, signal strength above a certain threshold is set as an interference source, and hotspots are identified based on the frequency of raptor appearances. The topological neural network prioritizes the shortest time path, and the output instruction set is standardized to one navigation point every 100 meters to ensure path smoothness.

[0014] The beneficial effects of this invention are as follows: 1. By integrating a micro biosensor with an external environment data receiving module, multi-dimensional biological signals and environmental data of homing pigeons are collected in real time. After processing by a signal fusion mechanism, a dynamic sensing dataset is formed, providing a comprehensive and accurate data foundation for subsequent analysis.

[0015] 2. Migratory bird and marine fish data are reconstructed into resolvable neural excitation signals for homing pigeons using knowledge transfer technology. Multiple candidate homing path fragments containing terrain avoidance strategies and energy consumption estimates are generated in a virtual geographic information system, providing multiple feasible path references for homing pigeons.

[0016] 3. By establishing a spatiotemporal convolution model using the historical homing trajectories of homing pigeons, common features of group flight decision-making are extracted. Combined with a micro barometer to sense the direction of micro airflow vortices in real time, the local heading angle of the path segment is dynamically corrected, and a continuous flight vector sequence resistant to turbulence interference is generated to improve the flight stability of homing pigeons.

[0017] 4. By using a chaotic convergence controller based on the Lyapunov exponent to analyze the stability of path segments, segments that meet the biological stability threshold are selected to ensure the stability of the homing pigeon's path and reduce flight risks.

[0018] 5. A topological neural network is used to connect the selected stable path segments end to end to synthesize the final homing navigation instruction set. At the same time, it avoids electromagnetic interference sources and predator activity hotspots, and outputs instructions that meet the principle of shortest time optimization, ensuring that the homing pigeons return to their homing nests safely and efficiently. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, 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 this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the AI ​​prediction and big data optimization system for homing routes of pigeon locators according to the present invention; Figure 2 This is a flowchart illustrating the AI ​​prediction and big data optimization method for homing routes in pigeon locators according to the present invention. Detailed Implementation

[0021] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0022] This invention proposes an AI prediction and big data optimization system and method for homing routes of pigeon locators.

[0023] like Figure 1 As shown, the system of this invention mainly includes: a miniature biosensor module, installed inside the pigeon locator, for real-time collection of pigeon flight physiological data; an external environment data receiving module, for acquiring meteorological and terrain data; a signal fusion module, for integrating physiological and environmental data to form a dynamic perception dataset; a knowledge transfer technology module, for reconstructing heterogeneous biological navigation data into neural excitation signals; a virtual geographic information system module, for generating candidate homing path segments based on neural excitation signals; a spatiotemporal convolution model module, for extracting group flight decision features based on the historical trajectory of the pigeon flock; a chaotic convergence controller module, for analyzing the stability of path segments; and a topological neural network module, for synthesizing stable segments into a continuous homing navigation instruction set.

[0024] like Figure 2 As shown, the method of the present invention mainly includes the following steps: Step 1: By integrating a micro biosensor with an external environmental data receiving module, the physiological tremor waveform, geomagnetic induction intensity, and visual spectral characteristics of the homing pigeon during flight are collected in real time. At the same time, atmospheric turbulence intensity and three-dimensional contours of terrain obstacles provided by meteorological satellites are acquired, ultimately forming a dynamic perception dataset that integrates biological instinctive behavior and geospatial changes.

[0025] Step 2: The celestial navigation neuron activation patterns of migratory birds and the geomagnetic memory maps of marine fish are reconstructed into resolvable neural excitation signals for homing pigeons through adversarial generative networks. Based on these signals, multiple candidate homing route fragments are generated in a virtual geographic information system. Each route fragment includes terrain avoidance strategies and energy consumption estimates.

[0026] Step 3: Establish a spatiotemporal convolution model of the historical homing trajectory of the pigeon flock and extract common features of the group's flight decision-making; use a miniature barometer to sense the direction of the microscopic airflow vortex at the location of the pigeon in real time, and combine the principles of atmospheric boundary layer fluid dynamics to dynamically correct the local heading angle of the path segment, generating a continuous flight vector sequence that is resistant to turbulence interference.

[0027] Step 4: Input the aforementioned path segments into the chaotic convergence controller, use the Lyapunov exponent to analyze the stability of each segment in phase space, and screen segments that meet the biological stability threshold; connect the segments end to end through a topological neural network, and output a continuous homing navigation instruction set that conforms to the shortest time optimal principle while avoiding electromagnetic interference sources and enemy activity hotspots.

[0028] The method of the present invention specifically includes the following steps: Step 1: By integrating a micro biosensor with an external environmental data receiving module, multi-dimensional biological signals and environmental data are collected in real time during the flight of the homing pigeon.

[0029] Specifically, a miniature triaxial accelerometer (designed to detect 0.01 mm displacement per second and output a three-dimensional acceleration sequence) is fixed inside the locator shell on the pigeon's back to collect physiological tremor waveforms in real time. A magnetometer employs the Hall effect principle, with a range covering 0.1 microtesla to 50 microtesla, measuring geomagnetic induction intensity to adapt to magnetic field variations in different geographical regions. A miniature spectrometer integrates narrowband filter technology to capture visual spectral characteristics at wavelengths from 500 nm to 700 nm, matching the pigeon's visual perception range and prioritizing the extraction of the green band. A meteorological satellite data receiving module is equipped with a dedicated loop antenna, and a low-frequency satellite signal module analyzes atmospheric turbulence intensity (unit: m / s) in real time and extracts the three-dimensional contours of terrain obstacles (point cloud structure, including spatial coordinates and elevation information). The atmospheric turbulence intensity threshold is set to above 5 m / s as a high-risk condition. The data acquisition frequency is strictly no less than 100 times per second; for example, when the pigeon accelerates or turns, changes in physiological tremor waveforms must be recorded in real time to avoid missing crucial biological signals. The signal fusion mechanism employs a time alignment algorithm, using interpolation compensation to eliminate device latency differences. For example, when satellite data is delayed by 0.01 seconds, the biosignal timestamp is automatically adjusted to ensure that the vibration waveform of the biosensor and the satellite terrain contour are precisely matched to the millisecond level within the same time window. The fusion process uses a moving average algorithm to align the time sequence, and the output dataset is updated once per second. This method is suitable for flat terrain and has low hardware costs, but its signal noise suppression capability under strong wind conditions needs to be verified. The fused dataset is stored in a structured list, with fields uniformly named timestamp (accurate to milliseconds), signal value (e.g., physiological tremor waveform amplitude, labeled unit), and spatial coordinates (latitude, longitude, and altitude). An automatic retransmission mechanism ensures data integrity, triggering a retransmission protocol when the signal loss rate exceeds 5%.

[0030] In the second embodiment, the accelerometer is replaced with a microelectromechanical system (MEMS) gyroscope, a dynamic range compensation algorithm is added, and the sampling rate is increased to 200 times per second; the magnetometer is replaced with a magnetoresistive sensor, the range is extended to 100 microtesla, and a temperature compensation module is embedded; the spectrometer adopts a multispectral array design (covering 300nm to 1000nm), and the visual feature extraction enhances the edge detection algorithm; the meteorological receiver module integrates redundant reception from multiple satellite sources, and a 3D reconstruction algorithm is added to the terrain contour analysis, i.e., generating a high-precision model based on point cloud interpolation; a Kalman filter is introduced into the fusion process to reduce signal jitter, and quality control markers, such as signal confidence scores ranging from 0 to 1, are added to the dataset fields. This variant is suitable for complex environments such as mountainous or urban areas, and performance drift needs to be tested within a temperature range of -10℃ to 40℃.

[0031] In the third embodiment, the accelerometer is simplified to a single-axis model, acquiring only vertical physiological tremor waveforms to reduce power consumption; the magnetometer and spectrometer are integrated into a single chip, with the measurement range reduced to 0.1-10 microtesla, and visual spectroscopy only analyzing the 550nm green band; meteorological data reception is changed to a preset regional meteorological database (not real-time satellite signals), and terrain contours use open-source geographic information system data; the fusion algorithm adopts a simple time-weighted average, and the dataset update frequency is reduced to once every 5 seconds. This variant sacrifices some accuracy but improves energy efficiency, is suitable for short-distance homing tasks, and requires verification of path matching in suburban environments, with a data alignment error of less than 5%.

[0032] Step 2 involves reconstructing the astronomical navigation neuron activation patterns (millisecond-level pulse sequences) of migratory birds and the geomagnetic field memory maps (rasterized matrices) of marine fish into resolvable neural excitation signals for homing pigeons using knowledge transfer technology. The amplitude is normalized to 0.5-5 volts, and multiple candidate homing path segments are generated in the virtual geographic information system. Each segment is no longer than 1 kilometer and includes terrain avoidance strategies and energy consumption estimates, with the unit uniformly set to joules.

[0033] The migration process requires prior verification of species physiological compatibility, such as a neuronal activation frequency matching rate higher than 80%. The implementation mechanism consists of three stages: First, heterogeneous data is extracted from a pre-constructed biological database. Second, the signal is reconstructed using a generative adversarial network: the generator is a three-layer fully connected neural network with 100 neurons in the input layer, 50 in the hidden layer, and 20 in the output layer; the discriminator is a convolutional neural network with a kernel size of 3×3. This three-layer neural network design ensures that the signal reconstruction error is less than 5%. Finally, in a virtual geographic information system (such as an open-source 3D globe platform), a standard terrain elevation dataset is used. Based on the reconstructed signal, a path generation algorithm is driven, employing a variant of the Dijkstra shortest path algorithm to automatically avoid steep slopes greater than 30 degrees. The energy estimation model uses bird flight dynamics formulas, outputting segment energy consumption values ​​based on the relationship between the average weight of a homing pigeon (300 grams) and its flight speed. This model is suitable for medium-resource equipment, with an air resistance coefficient set to 0.25. The robustness of knowledge transfer is ensured through adversarial training. The generative adversarial network consists of a generator and a discriminator. The generator takes migratory bird neuron patterns as input, with a sampling rate of 100 pulses per second, and outputs a reconstructed signal. The discriminator verifies that the signal conforms to the biological constraints of pigeons, such as the pulse interval being no more than 0.1 seconds.

[0034] In the second embodiment, a transfer learning layer is added to the generative adversarial network, pre-trained on a dataset of 10 bird species. The neuron patterns are expanded to a three-dimensional spatiotemporal model, including time, space, and frequency dimensions. Pulse width modulation technology is introduced to adapt to the neural sensitivity of pigeons in the reconstructed signal. The virtual system is upgraded to a real-time rendering engine, and path generation is changed to a genetic algorithm to generate diverse segments, such as adding detour path options. The terrain avoidance strategy incorporates the visual feature data from the first step, and the energy estimation adds an environmental wind speed correction coefficient, such as halving energy consumption when the wind is tailwinding. This variant needs to be verified for species compatibility through in vitro neural experiments.

[0035] In the third embodiment, the generative adversarial network is simplified to a lightweight single-layer generator, reducing the number of neurons to 50. The knowledge base is pre-compressed into a rule engine, such as a pulse frequency mapping table. The virtual system runs offline, path segment generation is batch-processed every 10 minutes, and energy estimation uses a linear approximation formula: energy consumption = speed. 2 × constant. This variant is suitable for low-power scenarios and requires testing the tolerance of reconstruction signal errors (set to 10%). After the input dataset in this step is verified for compatibility, the migration is started. The output format of the path segment is uniformly a coordinate sequence and a set of energy consumption labels. All embodiments share the output standard. For example, a biological matching test module is added before migration to ensure that the neural excitation signal will not trigger a stress response in the pigeon.

[0036] Step 3: Utilize historical homing trajectory data of pigeon flocks (over 10,000 records) to establish a spatiotemporal convolutional model, extract common features of flock flight decisions, such as a heat map of flock obstacle avoidance preference density, and use a miniature barometer (accuracy 0.1 Pascal, sampling vortex direction vectors 5 times per second) to perceive the microscopic airflow vortex direction of the pigeons' location in real time. Combine this with a simplified version of the Navier-Stokes equations to calculate the influence of pressure gradient force on heading, dynamically correct the local heading angle of the path segment (change not exceeding 10 degrees), and finally generate a continuous flight vector sequence resistant to turbulence interference. The output sequence fluctuates by less than 5 degrees under wind speed interference.

[0037] In the first embodiment, the baseline convolutional model is configured as a two-dimensional convolutional layer with a kernel size of 3×3. The input is a historical trajectory time series with a window length of 10 seconds, and the output feature map is a population density distribution. A miniature barometer is integrated into the side of the locator, and eddy current sensing data is processed once per second. The heading correction formula references the viscous fluid drag equation, updates once per second, and the output sequence format is an angle-velocity pair, such as a heading of 0 degrees due north and a velocity of 10 m / s. This embodiment is suitable for stable weather conditions, and the deviation rate needs to be verified to be less than 5% in simulations.

[0038] In the second embodiment, the convolutional model is extended to three dimensions, and an attention mechanism is added to feature extraction to focus on highly turbulent regions. The barometer is upgraded to a four-probe array, and turbulence intensity classification is added to vortex direction analysis, with weak vortex wind speeds less than 3 m / s and strong vortex wind speeds greater than 5 m / s. The correction formula adopts the Reynolds-averaged Navier-Stokes equations to dynamically calculate the pressure gradient force, and the heading angle correction step is refined to 0.5 degrees. The output sequence is fused with the energy estimate from the second step to achieve a balance between energy consumption and heading. This variant is suitable for mountainous or coastal highly turbulent areas and requires wind tunnel calibration.

[0039] In the third embodiment, the spatiotemporal convolutional model is replaced with a graph neural network to model individual pigeon interactions, such as the weighting coefficients of following behavior; barometer data and group characteristics are fused in real time, and a group voting mechanism is introduced for heading correction, adopting the direction favored by more than 60% of the individuals; eddy current modeling is simplified to one-dimensional vector calculation to reduce computational load. This variant is suitable for dense group flight, and the efficiency of group decision-making needs to be tested.

[0040] Step 3: After the path segment is input in Step 2, model feature extraction and real-time perception are performed in parallel. The corrected output is fed back to the path continuity. All embodiments include an anti-interference verification module, such as simulating eddy current intensity of 10 m / s to test path stability.

[0041] Step 4: Stable path segments from the flight vector sequence output in Step 3 are selected. This is done by analyzing segment stability using a chaotic convergence controller based on the Lyapunov exponent; an exponent less than 0.1 indicates stability. The selection threshold uses biological stability-oriented experimental data, such as a pigeon heart rate variation within 5%. A topological neural network (fully connected structure, heuristic search of connection rules) is then used to concatenate the segments end-to-end to synthesize the final homing navigation command set. This set includes coordinate sequences and timestamps, while avoiding electromagnetic interference sources (such as high-voltage line coordinates) and predator activity hotspots (based on raptor sighting frequencies). The topological neural network prioritizes the shortest time path, and the output command set is standardized to one navigation point every 100 meters.

[0042] As a first embodiment, the standard convergence controller design employs a phase space reconstruction method with an embedding dimension of 3, an exponential calculation window length of 5 seconds, and a stability threshold of 0.05. The topological neural network has two layers, and the connection algorithm is a variant of the A* search algorithm. The avoidance module sets a signal strength threshold of 10 microtesla or higher as an interference source, and generates the output instruction set once per second. This embodiment ensures path smoothness and is suitable for general scenarios.

[0043] In the second embodiment, the exponent calculation is optimized into an online estimation algorithm, with a sliding window updated every second, and the threshold dynamically adjusted, such as widening to 0.2 on rainy days; a reinforcement learning layer is added to the neural network, and the connection weights are optimized based on real-time energy consumption feedback; the avoidance strategy incorporates enemy reports monitored by the drone. This variant requires testing for robustness to the dynamic threshold.

[0044] In the third embodiment, exponential analysis is replaced with fragment variance evaluation, where stability is defined as a variance less than 0.1. The topology is simplified to nearest neighbor connection rules, avoiding the use of a static coordinate blacklist. This variant is suitable for low-computing-power devices and requires verification that the path synthesis time is less than 0.5 seconds.

[0045] After the input sequence undergoes stability screening in step 4, the neural network synthesizes the path and outputs the instruction set, which is then fed back to the locator for execution.

[0046] The beneficial effects of this invention are as follows: 1. By integrating a miniature biosensor with an external environment receiving module, the system collects real-time data on the physiological tremor characteristics, geomagnetic induction intensity, visual spectral characteristics, atmospheric turbulence, and terrain obstacle outlines of homing pigeons during flight. The data is then integrated using a time synchronization and averaging fusion algorithm to form a dynamic perception dataset, which significantly improves navigation accuracy and provides a reliable data foundation for homing pigeons to return home.

[0047] 2. By using knowledge transfer technology, the navigation neuron patterns of migratory birds and the geomagnetic memory maps of fish are reconstructed into resolvable neural excitation signals for homing pigeons. In a virtual geographic system, candidate homing path fragments containing terrain avoidance strategies and energy consumption estimates are generated, which effectively expands the diversity of path selection and optimizes the feasibility and adaptability of the homing process of homing pigeons.

[0048] 3. By establishing a spatiotemporal analysis model using the historical trajectories of homing pigeon flocks, the characteristics of group flight decision-making are extracted. At the same time, the direction of airflow vortex is monitored in real time through a miniature air pressure sensing device. Combined with simplified fluid dynamics equations, the local heading angle of the path is dynamically corrected to generate an interference-resistant continuous flight sequence, which significantly enhances the flight stability of homing pigeons in complex airflow environments and reduces path deviation.

[0049] 4. By analyzing the stability of path segments in mathematical space through a chaotic convergence controller, segments that meet the biological instinct steady-state threshold are selected, effectively ensuring the stability of the homing pigeon's path, significantly reducing the risk of flight deviation and energy consumption, and improving homing efficiency.

[0050] 5. By connecting the selected stable path segments end to end through a topological network, while avoiding electromagnetic interference and areas of predator activity, a continuous homing navigation instruction set is synthesized, realizing time-optimized path navigation and significantly improving the safety and efficiency of homing pigeons returning home.

[0051] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A system for AI prediction and big data optimization of homing routes for pigeon locators, characterized in that, include: A miniature biosensor module, installed inside the pigeon locator, is used to collect real-time physiological data of the pigeon's flight. External environment data receiving module, used to acquire meteorological and topographic data; The signal fusion module is used to integrate physiological and environmental data to form a dynamic sensing dataset; The knowledge transfer technology module is used to reconstruct heterogeneous biological navigation data into neural excitation signals; The virtual geographic information system module is used to generate candidate homing path fragments based on neural excitation signals; The spatiotemporal convolutional model module is used to extract group flight decision features based on the historical trajectories of homing pigeon flocks; Chaotic convergence controller module, used to analyze the stability of path segments; The topological neural network module is used to synthesize a continuous homing navigation instruction set from stable segments.

2. The homing route AI prediction and big data optimization system as described in claim 1, characterized in that, The signal fusion module uses a time alignment algorithm to eliminate device delay differences and applies a moving average algorithm to align timing, ensuring that biosignals and environmental data are accurately matched within a millisecond time window, and outputting a structured dataset updated once per second. The data fields include timestamps, signal values, and spatial coordinates. Data integrity is ensured through an automatic retransmission mechanism, which triggers retransmission when the signal loss rate exceeds a preset threshold.

3. The homing route AI prediction and big data optimization system as described in claim 1, characterized in that, The knowledge transfer technology module includes generating adversarial networks to reconstruct neural excitation signals. The generator is a three-layer fully connected neural network, and the discriminator is a convolutional neural network. The amplitude of the reconstructed signal is normalized and candidate path segments are generated through a virtual geographic information system. The length of each path segment includes terrain avoidance strategies and energy consumption estimates. The migration process pre-verifies species physiological compatibility to ensure that the neuron activation frequency matching degree is higher than a threshold.

4. The homing route AI prediction and big data optimization system as described in claim 1, characterized in that, The chaotic convergence controller module analyzes the stability of path segments in phase space based on the Lyapunov index and sets a biological stability threshold to screen stable segments; if the stability index is less than the preset value, it is considered to meet the requirements. The screening process ensures that the homing pigeons' homing path reduces flight risks.

5. The homing route AI prediction and big data optimization system as described in claim 1, characterized in that, The topological neural network module adopts a fully connected heuristic search rule to connect stable path segments end to end to generate a continuous homing navigation instruction set with a navigation point every 100 meters; at the same time, it avoids electromagnetic interference sources and predator activity hotspots, and the output instruction set conforms to the principle of shortest time optimization, ensuring that the homing pigeons return to their nests safely and efficiently.

6. A method for AI prediction and big data optimization of homing routes for pigeon locators, characterized in that, include: By integrating a micro biosensor with an external environment data receiving module, the system collects real-time data on pigeon physiological tremor waveforms, geomagnetic induction intensity, visual spectral characteristics, and atmospheric turbulence terrain contours. A time alignment algorithm is then used to fuse these data to form a dynamic sensing dataset. Through knowledge transfer technology, the activation patterns of celestial navigation neurons of migratory birds and the geomagnetic memory maps of marine fish are reconstructed into resolvable neural excitation signals for homing pigeons, and candidate homing path fragments including terrain avoidance strategies and energy consumption estimates are generated in a virtual geographic information system. A spatiotemporal convolutional model was established using the historical homing trajectories of homing pigeons to extract the group's flight decision-making features. Combined with a micro barometer to perceive the direction of micro airflow vortices in real time, the path segments were dynamically corrected to generate anti-turbulence interference flight vector sequences. By analyzing the stability of fragments based on the Lyapunov exponent using a chaotic convergence controller and selecting fragments that meet the biological homeostasis threshold, a topological neural network is used to connect the fragments end to end to synthesize a continuous homing navigation instruction set, while avoiding electromagnetic interference sources and enemy activity hotspots.

7. The homing route AI prediction and big data optimization method as described in claim 6, characterized in that, Real-time collection of homing pigeon physiological data includes using a miniature triaxial accelerometer to detect physiological tremor waveforms, a magnetometer to measure geomagnetic induction intensity, a spectrometer to capture visual spectral features, and a meteorological satellite data receiving module to analyze atmospheric turbulence intensity and terrain contours; the data acquisition frequency is no less than one hundred times per second, and real-time recording is performed when the homing pigeon accelerates or turns to avoid signal loss.

8. The homing route AI prediction and big data optimization method as described in claim 6, characterized in that, The reconstructed neural excitation signal is achieved through a generative adversarial network. The generator is input with migratory bird neuron patterns, and the discriminator verifies that the signal conforms to the biological constraints of homing pigeons. The reconstructed signal is used to drive the path generation algorithm in the virtual geographic information system, automatically avoid steep slope obstacles with a slope greater than a preset angle, and calculate the estimated energy consumption based on the bird flight dynamics formula.

9. The homing route AI prediction and big data optimization method as described in claim 6, characterized in that, The dynamic correction path segment includes: establishing a spatiotemporal convolution model to extract population density distribution features, sensing the vortex direction through a micro barometer, and calculating the heading correction using the viscous fluid resistance equation to generate an anti-turbulence interference flight vector sequence.

10. The homing route AI prediction and big data optimization method as described in claim 6, characterized in that, The method avoids electromagnetic interference sources and predator activity hotspots by setting a signal strength threshold above which is considered an interference source, and identifies hotspots based on the frequency of raptor appearances; the topological neural network prioritizes the shortest time path, and the output instruction set is standardized to one navigation point every 100 meters to ensure path smoothness.

Images