A method and system for the emergence of bionic swarm trajectories of drones based on spatiotemporal gradient fields

Abstract

This invention provides a method and system for the emergence of bionic swarm trajectories for drones using a spatiotemporal gradient field. This method relates to the technical field of drone swarm collaborative navigation and adaptive control. The method involves generating a four-dimensional spatiotemporal gradient field tensor matrix through a dynamic field-coupled convolutional network. A set of bionic swarm interaction rules is constructed through nonlinear differential game modeling, combining density-velocity response data of fish school escape behavior and drone dynamic constraints. Individual drone trajectory basis functions are determined using a spatiotemporal finite element discretization method combined with GPU-accelerated parallel solution. Swarm reorganization instructions are generated through Lyapunov function optimization processing. Finally, a bionic swarm trajectory emergence scheme for drones, enabled by the spatiotemporal gradient field, is generated by integrating the updated spatiotemporal gradient field, swarm reorganization instructions, and historical trajectory data. This invention provides drone swarms with high-precision, low-latency, and robust autonomous navigation capabilities, thereby improving mission success rates.

Description

Technical Field

[0001] The present invention relates to the technical field of cooperative navigation and adaptive control of drone swarms, and in particular to a method and system for bionic drone swarm trajectory emergence in a spatiotemporal gradient field. Background Art

[0002] The current UAV swarm collaborative navigation technology in complex environments mainly relies on centralized path planning algorithms and traditional bionic swarm intelligence models (such as Boids rules and particle swarm optimization). These technologies perform well in static or low-dynamic environments, but expose significant defects in complex dynamic scenes: 1. Traditional methods rely on static or low-frequency updated obstacle maps and cannot respond to dynamic expansion (such as the dynamic changes of ice cracks) and sudden interference (instantaneous wind speed > 15m / s) in real time. For example, the re-planning delay of the RRT algorithm in the crack expansion scenario is as high as 2-5 seconds, resulting in a significant increase in the risk of crash. In addition, the existing bionic model lacks a physical field coupling mechanism and cannot dynamically correct the trajectory through environmental forces (such as wind and ocean currents), resulting in a swarm disorder rate of over 65% under sudden wind disturbances; 2. Traditional multi-objective optimization (such as NSGA-II) does not take into account the decrease in motor efficiency caused by complex environments. The fixed weight distribution makes it difficult to balance energy consumption and obstacle avoidance. For example, in sea or ice environments, low temperature or high humidity conditions will affect the motor performance and Battery life; 3. Existing GPU acceleration solutions (such as CUDA uniform grid) do not address risk grading. Single-precision calculations across the entire area result in trajectory errors >0.5m in high-risk areas and overcalculation in safe areas. Furthermore, in sea or icy environments, drones need to cope with dynamically changing ocean currents and wind fields, and existing technologies cannot effectively adapt to these dynamic environments. For example, when the environment suddenly changes, full path replanning is required, and a swarm of thousands of drones takes >10 minutes, seriously lagging behind mission requirements. 4. In icy or sea environments, collaborative navigation of drone swarms also faces special challenges. For example, sea waves and currents significantly affect the flight stability of drones, and existing navigation technologies lack effective modeling and compensation for these factors. Summary of the Invention

[0003] The purpose of the present invention is to provide a method and system for the emergence of bionic swarm trajectories of drones in a spatiotemporal gradient field to improve the above-mentioned problems. To achieve the above-mentioned purpose, the technical solutions adopted by the present invention are as follows:

[0004] In a first aspect, the present application provides a method for the emergence of bionic swarm trajectories of drones in a spatiotemporal gradient field, comprising:

[0005] Obtain satellite remote sensing ice crack distribution data, LiDAR real-time terrain scanning data, ocean current velocity field data, surface temperature gradient data and instantaneous wind speed vector data, and generate a four-dimensional spatiotemporal gradient field tensor matrix through a dynamic field coupled convolutional network;

[0006] Based on the four-dimensional space-time gradient field tensor matrix, combined with the density-velocity response data of fish escape behavior and the dynamic constraints of the UAV, a set of bionic swarm interaction rules is constructed through nonlinear differential game modeling.

[0007] Using the bionic swarm interaction rule set and the four-dimensional space-time gradient field tensor matrix, the space-time finite element discretization method is adopted and combined with GPU accelerated parallel solution to determine the individual trajectory basis functions of the UAV;

[0008] Based on the real-time update data of the drone's individual trajectory basis functions and environmental sensors, the four-dimensional space-time gradient field tensor matrix is ​​dynamically adjusted through Lyapunov function optimization processing to obtain the updated space-time gradient field and generate group reorganization instructions;

[0009] By integrating the updated spatiotemporal gradient field, group reorganization instructions and historical trajectory data, and applying the quantum annealing-NSGAIII hybrid optimization algorithm, a bionic group trajectory emergence scheme for drones enabled by the spatiotemporal gradient field is generated.

[0010] Preferably, the acquisition of satellite remote sensing ice crack distribution data, LiDAR real-time terrain scanning data, ocean current velocity field data, surface temperature gradient data and instantaneous wind speed vector data generates a four-dimensional spatiotemporal gradient field tensor matrix through a dynamic field coupled convolutional network, which includes:

[0011] Ice surface images are acquired using synthetic aperture radars carried by polar-orbiting satellites. A crack edge detection algorithm is used to extract binary crack distribution maps with a spatial resolution of ≤1m, resulting in satellite remote sensing ice crack distribution data. Raw point cloud data is collected using a solid-state lidar carried by an unmanned aerial vehicle (UAV). Multi-frame point clouds are registered using a point cloud registration algorithm to generate a digital elevation model, resulting in real-time LiDAR terrain scanning data. Doppler velocity meters built into a 500m-spaced terrain sensor array measure horizontal flow velocity vectors in real time, and ocean current velocity field data are generated using Kriging spatial interpolation. A high-resolution infrared thermal imager carried by the UAV acquires the ice surface temperature distribution matrix at a sampling rate of 5Hz, and the spatial gradient field is calculated using the Sobel operator to obtain surface temperature gradient data. Instantaneous wind speed vector data is obtained by performing Kriging interpolation on the three-dimensional wind speeds obtained from a network of complex weather stations.

[0012] Satellite remote sensing ice crack distribution data, LiDAR real-time terrain scanning data, ocean current velocity field data, surface temperature gradient data and instantaneous wind speed vector data are encoded into a five-channel space-time tensor, and the five-channel space-time tensor is used as the input of the dynamic convolutional network to construct a dynamic convolutional neural network. Its convolution kernel weights are jointly modulated by the ice crack curvature and temperature gradient, and the output is a coupled drift field vector. Combined with the coordinates and priority weights provided by the mission requirement system, a four-dimensional space-time gradient field tensor matrix is ​​generated, where the four-dimensional space-time gradient field tensor matrix includes the gravitational field, repulsive field and drift field.

[0013] Preferably, the four-dimensional space-time gradient field tensor matrix is ​​combined with the density-velocity response data of the fish escape behavior and the UAV dynamic constraints to construct a bionic group interaction rule set through nonlinear differential game modeling, which includes:

[0014] Based on the group density-speed observation data of fish escape behavior, a Sigmoid function is used to fit the nonlinear mapping relationship between density and speed to generate a bionic density-speed response constraint curve;

[0015] The bionic density-velocity response constraint curve is combined with the preset UAV dynamics model to construct a Hamiltonian function and generate a dynamic feasible solution space for the nonlinear differential game. The UAV dynamics model is obtained by integrating polar parameter measurements, gradient field coupling, and bionic rule constraints.

[0016] Based on the ice crack repulsion gradient field component in the four-dimensional space-time gradient field tensor matrix, the real-time relative position of the drone group is integrated, and the inverse proportional repulsion potential energy superposition algorithm is used to generate the group obstacle avoidance and collision avoidance joint force vector field;

[0017] Based on the data of the dynamic feasible solution space, the combined force vector field of swarm obstacle avoidance and collision avoidance, and the drift field, a set of bionic swarm interaction rules is obtained through Bourns-Nash equilibrium solution and heading collaborative optimization, which includes static constraints and dynamic collaborative strategies.

[0018] Preferably, the ice crack repulsion gradient field component in the four-dimensional space-time gradient field tensor matrix is ​​integrated with the real-time relative position of the drone group, and an inverse proportional repulsion potential energy superposition algorithm is used to generate a group obstacle avoidance and collision avoidance joint force vector field, which includes:

[0019] Based on the ice crack repulsion gradient field component in the four-dimensional space-time gradient field tensor matrix, the ice crack repulsion gradient field covering the global mission area is calculated;

[0020] The local repulsive gradient in the ice crack repulsive gradient field is extracted and combined with the real-time position of the UAV to generate the anti-collision force between groups through the inverse proportional potential energy superposition;

[0021] The local repulsive force gradient and the vector of the anti-collision force between the groups are superimposed to obtain the total obstacle avoidance and collision avoidance combined force of the individual drones.

[0022] Preferably, the bionic swarm interaction rule set and the four-dimensional space-time gradient field tensor matrix are used to determine the individual trajectory basis functions of the drones by adopting the space-time finite element discretization method and combining GPU accelerated parallel solution, which includes:

[0023] Based on the four-dimensional space-time gradient field tensor matrix and the bionic swarm heading coordination constraint, a Fourier-Legendre joint basis function expansion method is used to construct a space-time threat perception basis function set. Its spatial wavenumber component is modulated by the ice crack distribution density, and its temporal frequency component is associated with the drift field time-varying rate. The heading coordination angle is embedded in the basis function phase through a cosine modulation term.

[0024] The spatiotemporal threat perception basis function set and the UAV dynamics model are discretized using the Galerkin weighted residual method to generate a trajectory coupled algebraic system. , where the coefficient matrix It is composed of the inner product of basis functions, and the right-hand term Contains orthogonal projections of group collision avoidance forces;

[0025] Based on a trajectory-coupled algebraic system and a four-dimensional space-time gradient field tensor matrix, a GPU-parallel solution for ice crevasse risk classification was implemented to obtain a risk zoning strategy. The strategy includes allocating 32×32 thread blocks to a high-risk zone within 0.5 m of the crevasse edge, a 1-2 m area outside the crevasse, a 64×64 thread block to a medium-risk zone and a flat ice surface, and a 128×128 thread block to a safe zone. Double-precision floating-point operations and preconditioning acceleration are used in the high-risk zone, while single-precision mixed iterative calculations are used in the medium- and low-risk zones.

[0026] The basis function coefficient vector is output through the risk partitioning strategy to generate the UAV continuous trajectory function.

[0027] Specifically, the quantum annealing-NSGAIII hybrid optimization algorithm is applied to the integrated updated spatiotemporal gradient field, group reorganization instructions and historical trajectory data to generate a spatiotemporal gradient field-enabled drone biomimetic group trajectory emergence scheme, which includes:

[0028] Based on the updated spatiotemporal gradient field, swarm reorganization instructions, and historical trajectory data, a multi-objective problem was constructed, including minimizing energy consumption, minimizing the number of ice crevasse collisions, and optimizing mission synchronization. The control input was constrained by the maximum thrust and bionic speed constraints of the UAV, ice crevasse collision determination was triggered by the curvature field threshold, and mission synchronization was measured by the arrival time difference.

[0029] The initial population is generated using a quantum annealing mechanism. Its Hamiltonian is designed to couple historical trajectory conflict statistics with the alignment of the repulsive field gradient direction. Each individual encodes the control input sequence and heading strategy of the drone. The quantum bit state reflects the decision variable. The coupling strength is determined by the swarm's collaborative risk quantification. The local field strength is correlated with the current gradient field direction consistency.

[0030] The population is optimized through a hybrid strategy of quantum tunneling mutation and dynamic reference point adjustment, and the final output Pareto optimal solution set is converted into a continuous trajectory scheme enabled by the spatiotemporal gradient field. Its coefficients are directly mapped by the optimization variables, and the heading instructions and reorganization marks are inherited from the decision coding of the optimal solution.

[0031] In a second aspect, the present application also provides a UAV bionic group trajectory emergence system with a spatiotemporal gradient field, comprising:

[0032] Acquisition module: used to obtain satellite remote sensing ice crack distribution data, LiDAR real-time terrain scanning data, ocean current velocity field data, surface temperature gradient data and instantaneous wind speed vector data, and generate a four-dimensional space-time gradient field tensor matrix through a dynamic field coupled convolutional network;

[0033] Building module: Used to construct a set of biomimetic swarm interaction rules through nonlinear differential game modeling based on the four-dimensional space-time gradient field tensor matrix, combined with the density-velocity response data of fish escape behavior and the dynamic constraints of the drone;

[0034] Solution module: Used to determine the individual trajectory basis functions of UAVs using the bionic swarm interaction rule set and the four-dimensional space-time gradient field tensor matrix, using the space-time finite element discretization method and GPU-accelerated parallel solution;

[0035] Processing module: used to update data in real time based on the individual trajectory basis functions of drones and environmental sensors, dynamically adjust the four-dimensional space-time gradient field tensor matrix through Lyapunov function optimization processing, obtain the updated space-time gradient field and generate group reorganization instructions;

[0036] Generation module: It is used to integrate the updated spatiotemporal gradient field, group reorganization instructions and historical trajectory data, and apply the quantum annealing-NSGAIII hybrid optimization algorithm to generate a bionic group trajectory emergence solution for drones enabled by the spatiotemporal gradient field.

[0037] In a third aspect, the present application also provides a drone bionic group trajectory emergence device with a spatiotemporal gradient field, comprising:

[0038] memory for storing computer programs;

[0039] A processor is used to implement the steps of the method for emerging the trajectory of a bionic swarm of drones in a spatiotemporal gradient field when executing the computer program.

[0040] In a fourth aspect, the present application also provides a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned method for the emergence of bionic swarm trajectories of drones based on a spatiotemporal gradient field.

[0041] The beneficial effects of the present invention are:

[0042] The present invention uses real-time fusion of satellite, LiDAR, and buoy array data using a four-dimensional space-time gradient field tensor matrix (repulsive field, gravitational field, and drift field) to construct a dynamic weighted model of the repulsive gradient at the edge of ice cracks, enabling millimeter-level perception of environmental threats. Biomimetic constraints are designed based on the density-velocity response curve of fish escape behavior, and Hamiltonian function optimization and space-time basis function expansion are combined to generate distributed collaboration rules with low communication dependency. Quantum annealing is used to initialize the population and NSGA-III multi-target search. Through risk-graded GPU parallelism (double-precision calculation of 32×32 thread blocks in high-risk areas) and dynamic reference point adjustment, the optimization time for a swarm of thousands of drones is compressed to within 5ms while ensuring trajectory accuracy in high-risk areas (≤0.1m), thus overcoming the technical difficulties of autonomous collaboration of drone swarms in dynamic polar environments.

[0043] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or understood by practicing the embodiments of the present invention. The purposes and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1 This is a flow chart of the method for bionic swarm trajectory emergence of drones using a spatiotemporal gradient field according to an embodiment of the present invention;

[0046] Figure 2 Schematic diagram of the structure of the UAV bionic swarm trajectory emergence system of the spatiotemporal gradient field described in an embodiment of the present invention;

[0047] Figure 3 Schematic diagram of the structure of the drone bionic swarm trajectory emergence device with spatiotemporal gradient field described in an embodiment of the present invention.

[0048] In the figure: 701, acquisition module; 702, construction module; 703, solution module; 704, processing module; 705, generation module; 800, drone bionic group trajectory emergence device with spatiotemporal gradient field; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0050] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are used only to distinguish the description and should not be understood as indicating or implying relative importance.

[0051] Example 1:

[0052] This embodiment provides a method for the emergence of bionic swarm trajectories of drones in a spatiotemporal gradient field.

[0053] See also Figure 1 , the figure shows that the method includes step S100, step S200, step S300, step S400 and step S500.

[0054] S100, obtains satellite remote sensing ice crack distribution data, LiDAR real-time terrain scanning data, ocean current velocity field data, surface temperature gradient data and instantaneous wind speed vector data, and generates a four-dimensional space-time gradient field tensor matrix through a dynamic field coupled convolutional network.

[0055] It can be understood that step S100 includes steps S101 and S102, wherein:

[0056] S101. Acquire ice surface images using synthetic aperture radars carried by polar-orbiting satellites, extract crack distribution binary maps with a spatial resolution of ≤1m using an ice crack edge detection algorithm, and obtain satellite remote sensing ice crack distribution data. Collect raw point cloud data using a solid-state lidar carried by an unmanned aerial vehicle (UAV), register multiple point clouds using a point cloud registration algorithm, generate a digital elevation model, and obtain LiDAR real-time terrain scanning data. Measure horizontal flow velocity vectors in real time using Doppler velocity meters built into a terrain sensor array with a spacing of 500m, and generate ocean current velocity field data through Kriging spatial interpolation. Obtain ice surface temperature distribution matrices using a high-resolution infrared thermal imager carried by the UAV at a sampling rate of 5Hz, and calculate the spatial gradient field based on the Sobel operator to obtain surface temperature gradient data. Obtain instantaneous wind speed vector data by performing Kriging interpolation on the three-dimensional wind speeds obtained from a complex environment weather station network.

[0057] It should be noted that the satellite remote sensing ice crack distribution data: ice surface images are obtained through synthetic aperture radar (SAR) carried by polar-orbiting satellites, and crack distribution binary maps are extracted through ice crack edge detection algorithms (based on Canny operator and morphological closing operation), with a spatial resolution of ≤1m and updated daily; UAVs are equipped with solid-state lidar (wavelength 905nm, scanning frequency 10Hz), and digital elevation models (DEMs) are generated through point cloud registration algorithms with an accuracy of ±0.1m; a terrain sensor array is deployed (spacing 500m), and a built-in Doppler flow meter is used to measure horizontal flow velocity vectors , sampling rate 1Hz; the drone is equipped with an infrared thermal imager (band 8-14μm, thermal sensitivity ≤0.05℃) to collect temperature distribution T ( x , y , t ), calculate the gradient field ; Complex environment weather station network (node ​​spacing 1km) provides three-dimensional wind speed , and a continuous wind field is generated through Kriging interpolation.

[0058] S102. Encode satellite remote sensing ice crack distribution data, LiDAR real-time terrain scanning data, ocean current velocity field data, surface temperature gradient data and instantaneous wind speed vector data into a five-channel space-time tensor, and use the five-channel space-time tensor as the input of a dynamic convolutional network, thereby constructing a dynamic convolutional neural network, whose convolution kernel weights are jointly modulated by the ice crack curvature and the temperature gradient, and the output is a coupled drift field vector. Combined with the coordinates and priority weights provided by the task requirement system, a four-dimensional space-time gradient field tensor matrix is ​​generated, wherein the four-dimensional space-time gradient field tensor matrix includes a gravitational field, a repulsive field and a drift field.

[0059] It should be noted that the above data is encoded into a five-channel tensor ,in, is a binary map of ice crack distribution (0 / 1 represents safe / dangerous areas); is the LiDAR elevation matrix; is the normalized physical field scalar or vector, and then the ice crack curvature-temperature gradient joint modulation mechanism is designed; wherein the convolution kernel is designed as: ,in σ is the Sigmoid activation function, W is a trainable weight matrix that outputs a dynamic coupling field Fdynamic ( x , y , t ). Among them, gravity field: the target point priority weight generates a potential energy gradient, that is, according to the mission material delivery target point and its priority, a time-varying gravity field is generated; repulsion field: the ice crack edge gradient (calculated based on the curvature of the LiDAR point cloud, the repulsion peak is generated in the area where the curvature > threshold); drift field: the ocean current and wind speed vector are superimposed to generate a time-varying drift correction term (dynamic Bayesian network predicts the field intensity distribution in the next 10 seconds), that is, through three layers of convolution (kernel size 3×3, step size 1) to extract the multi-source field coupling characteristics and output the dynamic drift field. The repulsion gradient, drift field and gravity gradient are integrated, that is, the formula is:

[0060]

[0061] in, is the space-time gradient field tensor, is the potential field gradient, pointing to the direction of the fastest descent of the combined potential field (repulsion + attraction), For the drift field.

[0062] in, The shortcomings of the traditional binary ice crack map are corrected by DEM elevation to avoid the misjudgment of the safety height of the drone on the slope terrain, and the DCNN convolution kernel weight is determined by the temperature gradient ∇ T Modulation, early response to ocean current mutations caused by ice melting, and the spatiotemporal gradient field G It is both the input of the differential game and the reverse optimization of its own parameters through trajectory data to form an adaptive control loop. Therefore, after obtaining the four-dimensional space-time gradient field tensor matrix, it provides a physical driving field for bionic swarm rule modeling.

[0063] S200, based on the four-dimensional space-time gradient field tensor matrix, combined with the density-velocity response data of fish escape behavior and the dynamic constraints of drones, a set of bionic swarm interaction rules is constructed through nonlinear differential game modeling.

[0064] It can be understood that step S200 includes steps S201, S202, S203, and S204, wherein:

[0065] S201. Based on the density-velocity observation data of the fish escape behavior, a Sigmoid function is used to fit the nonlinear mapping relationship between density and velocity to generate a bionic density-velocity response constraint curve. The calculation formula is as follows:

[0066]

[0067] Where, is the density-velocity response constraint curve, is the maximum allowed speed of the drone, is the drone population density, is the critical density threshold, k is the steepness coefficient of the Sigmoid curve, e is the base of the natural logarithm, ;

[0068] It should be noted that when a school of fish encounters a predator, its escape velocity exhibits a nonlinear relationship with the density of the group (the higher the density, the lower the upper limit of the individual speed to avoid collision). This biological behavior law is quantified as a mathematical constraint and embedded in the UAV differential game model to form a bionic differential constraint. The curve is an inequality constraint of the differential game, which limits the maximum speed of the UAV in a dense group and prevents collision or loss of control due to excessive speed during low-altitude flight in the polar regions. Among them, the peak escape velocity of the fish school is ≤1.5m / s, and the fitting of the fish school data yields k =3.0, make sure ρ ∈[1.8,2.2]kg / m3, the velocity decays rapidly.

[0069] S202. The bionic density-velocity response constraint curve is combined with the preset UAV dynamics model to construct a Hamiltonian function, generating a dynamic feasible solution space for the nonlinear differential game. The UAV dynamics model is obtained by integrating polar parameter measurements, gradient field coupling, and bionic rule constraints. The calculation formula for constructing the Hamiltonian function is as follows:

[0070]

[0071] Where, For drone control input, is the Hamiltonian function, is the co-state vector, T is the matrix, is the state equation of the UAV, is the control input matrix;

[0072] It should be noted that the construction of the UAV dynamics model includes: extracting from the existing UAV dynamics theory to describe the physical relationship between position, speed, and attitude, and limiting the control input by measuring the maximum thrust at 30°C from 40N to 35N, and measuring the maximum thrust at 30°C through polar tests. μ=0.03 (only 1 / 10 of the conventional ground), the attitude control parameters need to be adjusted, the four-dimensional space-time gradient field tensor matrix obtained above is embedded in the dynamic equation, the environmental force coupling is obtained, and the speed is limited by the fish school rule. In the Hamiltonian function, the state equation f ( x ) contains the gradient field and drift field , the control input satisfies , we obtain the dynamic feasible solution space of the nonlinear differential game and ensure that the trajectory meets the physical limits of the UAV.

[0073] S203, based on the ice crack repulsion gradient field component in the four-dimensional space-time gradient field tensor matrix, integrating the real-time relative position of the drone group, and using the inverse proportional repulsion potential energy superposition algorithm, generate a swarm obstacle avoidance and collision avoidance joint force vector field;

[0074] It should be noted that this step is to ensure that drones can avoid ice cracks and collisions in polar environments.

[0075] In this embodiment, step S203 includes S2031, S2032, and S2033, wherein:

[0076] S2031. Calculate the ice crack repulsion gradient field covering the global task area based on the ice crack repulsion gradient field component in the four-dimensional space-time gradient field tensor matrix;

[0077] S2032. Extract the local repulsive force gradient in the ice crack repulsive force gradient field, combine it with the real-time position of the UAV, and generate the anti-collision force between groups by superposition of inversely proportional potential energy. The calculation formula is as follows:

[0078]

[0079] Where, 、 denote the positions of drones i and j respectively, γ =0.8 is the repulsive force weight coefficient, For drones i The inter-group collision avoidance force, Euclidean distance between two drones;

[0080] S2033. Superimpose the local repulsive force gradient and the anti-collision force between the groups to obtain the total obstacle avoidance and collision avoidance combined force of the individual drones. The calculation formula is as follows:

[0081]

[0082] Where, is the local repulsive gradient, To prevent collisions between groups, is the combined force vector field of group obstacle avoidance and collision avoidance.

[0083] It should be noted that the repulsive weight γ Adaptively adjust with the ice crevasse curvature to ensure the avoidance priority of high-risk areas (such as the edge of the ice crevasse). In this step, the edge curvature of the ice crevasse Used to dynamically adjust the repulsive force strength: >0.7 (high-risk fissure area), repulsion weight γ Automatic enhancement (such as γ →1.2 γ ), forcing the drone to make a larger detour; when ≤0.7 (low-risk area), using standard weight γ =0.8, balancing obstacle avoidance and flight efficiency. The final result is Fsep It is a vector force whose direction is determined by the repulsive force gradient of the ice crack and the repulsive force between drones, ensuring that drones can achieve autonomous obstacle avoidance and group collaboration in the complex polar environment, and enable drones to intelligently identify and prioritize avoiding high-risk crack areas during flight, while maintaining a safe distance between groups.

[0084] S204. Based on the data of the dynamic feasible solution space, the combined force vector field of the swarm obstacle avoidance and collision avoidance, and the drift field, a set of bionic swarm interaction rules is obtained through the Burkholderia Nash equilibrium solution and heading collaborative optimization, which includes static constraints and dynamic collaborative strategies.

[0085] It should be noted that the space of dynamically feasible solutions includes a set of candidate trajectories that satisfy polar dynamics constraints (such as low-temperature thrust limits and ice surface friction coefficients), and each trajectory is accompanied by multi-objective evaluation indicators such as energy consumption, time, and safety; a group obstacle avoidance and collision avoidance combined force vector field, in which the force direction of each point is dominated by the ice crack repulsion gradient, and the intensity is dynamically adjusted with the curvature (the intensity in high-risk areas is increased by 50%); and the drift field data includes the spatiotemporal changes of ocean currents and wind speeds, and its gradient can predict the environmental disturbance trend in the next 10 seconds.

[0086] It is understandable that each drone acts as an independent player, and its strategy space is a subset of the dynamically feasible solutions (e.g., only retaining the top 20% of candidate trajectories in terms of energy consumption).

[0087] Therefore, the main objectives are: Then, weights are assigned and an asynchronous distributed ADMM algorithm is used. Each drone iteratively optimizes its strategy based on neighbor information: that is, it adjusts candidate trajectories based on the current joint force; exchanges trajectory scores through polar low-bandwidth channels (such as LoRa) and negotiates conflicting segments; when the strategy change rate of all drones is less than 1%, it is determined to be a Nash equilibrium. (Based on the direction of the gradient field); if it is detected , Brownian motion compensation is activated, and the heading is subject to stochastic differential equations. After the wind disturbance ends, coordination is quickly restored through the Lyapunov function. Therefore, through drift field gradient prediction and local game theory, an 80% task completion rate is maintained even with a 30% packet loss rate. Distributed solving reduces the computational load to one-fifth of that of centralized methods. The density-velocity relationship of fish escape is encoded as a constraint. Drift field gradients replace biochemical signaling to achieve physically driven coordination.

[0088] S300, using the bionic group interaction rule set and the four-dimensional space-time gradient field tensor matrix, adopts the space-time finite element discretization method and combines GPU accelerated parallel solution to determine the individual trajectory basis function of the UAV.

[0089] It can be understood that step S300 includes S301, S302, S303 and S304, wherein:

[0090] S301. Based on the four-dimensional space-time gradient field tensor matrix and the bionic swarm heading coordination constraint, a Fourier-Legendre joint basis function expansion method is used to construct a space-time threat perception basis function set. Its spatial wavenumber component is modulated by the ice crack distribution density, and its temporal frequency component is associated with the drift field time-varying rate. The heading coordination angle is embedded in the basis function phase through the cosine modulation term. Its calculation formula is as follows:

[0091]

[0092] Where, For the k spatiotemporal coupling basis functions, is the spatial fluctuation term, which characterizes the periodic change of the basis function in the horizontal plane. is the spatial wave number modulated by the ice crack distribution, is the time frequency parameter, is the group heading coordination angle, is the space-time gradient field tensor G The modulus length reflects the intensity of environmental threats. is the group heading coordination angle, t is the time variable;

[0093] S302, discretize the spatiotemporal threat perception basis function set and the UAV dynamics model through the Galerkin weighted residual method to generate a trajectory coupling algebraic system , where the coefficient matrix It is composed of the inner product of basis functions, and the right-hand term Contains orthogonal projections of group collision avoidance forces;

[0094]

[0095] Where, is the integral operation in the space-time domain, For drones i The partial derivative of the trajectory function with respect to time t, is the negative gradient of the ice crack repulsive field, is the environmental drift field, For the control input of the drone, is the kth space-time basis function, is the product of the infinitesimal elements of space and time;

[0096] S303. Based on the trajectory coupled algebraic system and the four-dimensional space-time gradient field tensor matrix, a GPU parallel solution process is implemented to obtain a risk zoning strategy. The risk zoning strategy includes allocating 32×32 thread blocks to a high-risk zone within 0.5 m of the ice crevasse edge, a 1-2 m area outside the crevasse, a 64×64 thread block to a medium-risk zone and a flat ice surface area, and a 128×128 thread block to a safe zone. Double-precision floating-point operations and preconditioning acceleration are used in the high-risk zone, while single-precision mixed iterative calculations are used in the medium- and low-risk zones.

[0097] S304: Output the basis function coefficient vector through the risk partitioning strategy to generate the UAV continuous trajectory function.

[0098] It should be noted that the computing resource priority is dynamically divided according to the ice crack curvature, and the computing density in the high-risk area is increased by 3 times; when the dynamic residual convergence control is performed, the real-time drift field change rate and the iterative residual sequence are input. When the sudden disturbance mode is in effect, the real-time drift field change rate is greater than

[0099] When the residual threshold is tightened to , sacrificing accuracy to ensure real-time performance; when activating the "catch-up-relaxation" iterative strategy, the previous step solution is used to accelerate convergence. In other words, when a sudden change in the drift field is detected, the residual threshold is tightened to and activate catch-up relaxation iterations. Understandably, in special circumstances, such as responding to sudden polar environmental events (e.g., snowstorms), dynamic threshold switching is used to balance computational accuracy with real-time performance. Skipping risk zoning and proceeding directly to homogenization can increase trajectory errors in high-risk areas by a factor of 2.3. Ignoring the repulsive gradient in the preconditioner increases the number of iterations at the ice crevasse edge by 70%. Staticizing the convergence strategy can lead to a solution failure rate exceeding 30% under sudden disturbances.

[0100] S400, based on the real-time update data of the drone's individual trajectory basis functions and environmental sensors, dynamically adjusts the four-dimensional space-time gradient field tensor matrix through Lyapunov function optimization processing, obtains the updated space-time gradient field and generates group reorganization instructions.

[0101] It is understandable that the calculation formula for the Lyapunov function optimization process in step S400 is as follows:

[0102]

[0103] Where, L ( t ) is a scalar function that characterizes the overall trajectory tracking error of the UAV swarm, is the actual position of UAV i, is the predicted position of UAV i, is the actual speed of UAV i, is the predicted speed of UAV i, λ is the speed error weight coefficient, and N is the size of the UAV swarm.

[0104] It should be noted that the process of "dynamically adjusting the four-dimensional space-time gradient field tensor matrix to obtain an updated space-time gradient field and generate group reorganization instructions" includes constructing the Lyapunov function, obtaining the group's overall stability index, and using the quasi-Newton method to reversely optimize the gradient field parameters, namely:

[0105]

[0106] Where, By calculating the chain rule, the trajectory error is related to the repulsive gradient If a sudden wind disturbance is detected , add eddy damping term to the drift field, and output the updated spatiotemporal gradient field G ′( x , y , z , t ); and according to the updated spatiotemporal gradient field and group connectivity index, when the connectivity index is greater than 0.7, the task area is re-divided based on the Voronoi diagram to generate subgroup target points. The subgroup target points are generated by the repulsive gradient drive and combined with the drift field correction amount Generate a reorganization instruction set, which includes the forced alignment of the heading of UAVs in high-risk areas and the forced alignment of UAVs in low-risk areas. Drift field corrects heading.

[0107] S500 integrates the updated spatiotemporal gradient field, group reorganization instructions and historical trajectory data, and applies the quantum annealing-NSGAIII hybrid optimization algorithm to generate a bionic group trajectory emergence solution for drones enabled by the spatiotemporal gradient field.

[0108] It can be understood that step S500 includes S501, S502 and S503, wherein:

[0109] S501. Based on the updated spatiotemporal gradient field, swarm reorganization instructions, and historical trajectory data, a multi-objective problem is constructed, including minimizing energy consumption, minimizing the number of ice crevasse collisions, and optimizing mission synchronization. The control input is constrained by the maximum thrust and bionic speed of the UAV, ice crevasse collision determination is triggered by the curvature field threshold, and mission synchronization is measured by the arrival time difference.

[0110] S502. Generate an initial population using a quantum annealing mechanism. Its Hamiltonian is designed to couple historical trajectory conflict statistics with the alignment of the repulsive field gradient direction. Each individual encodes the control input sequence and heading strategy of the drone. The quantum bit state reflects the decision variable. The coupling strength is determined by the quantization of the group's collaborative risk. The local field strength is correlated with the current gradient field direction consistency.

[0111] It should be noted that the quantum annealing mechanism is used to generate the initial population, and its Hamiltonian is designed as:

[0112]

[0113] Where, is the quantum bit state, is the coupling strength, The local field strength is positively correlated with the alignment of the repulsive gradient direction.

[0114] S503. Optimize the population through a hybrid strategy of quantum tunneling mutation and dynamic reference point adjustment. The final output Pareto optimal solution set is converted into a continuous trajectory scheme enabled by the spatiotemporal gradient field. Its coefficients are directly mapped by the optimization variables. The heading instructions and reorganization marks are inherited from the decision code of the optimal solution. The calculation formula is as follows:

[0115]

[0116] Where, is the optimized UAV trajectory function, K is the number of basis functions, is the optimal basis function coefficient, is the horizontal position x and the mission time t, Space-time basis functions.

[0117] The present invention uses a four-dimensional space-time gradient field tensor matrix to fuse satellite, LiDAR, and buoy array data in real time to construct a dynamic weighted model of the repulsive gradient at the edge of ice cracks, achieving millimeter-level perception of environmental threats. Based on the density-velocity response curve of fish escape behavior, bionic constraints are designed, and Hamiltonian function optimization and space-time basis function expansion are combined to generate distributed collaboration rules with low communication dependence. Quantum annealing is used to initialize the population and NSGA-III multi-target search. Through risk-graded GPU parallelism and dynamic reference point adjustment, the optimization time of a thousand-aircraft group is compressed to within 5ms, overcoming the technical difficulties of autonomous collaboration of UAV groups in dynamic polar environments.

[0118] Example 2:

[0119] like Figure 2 As shown, this embodiment provides a UAV bionic group trajectory emergence system with a spatiotemporal gradient field, see Figure 2 The system comprises:

[0120] Acquisition module 701: used to acquire satellite remote sensing ice crack distribution data, LiDAR real-time terrain scanning data, ocean current velocity field data, surface temperature gradient data and instantaneous wind speed vector data, and generate a four-dimensional space-time gradient field tensor matrix through a dynamic field coupled convolutional network;

[0121] Construction module 702: for constructing a bionic swarm interaction rule set through nonlinear differential game modeling based on the four-dimensional space-time gradient field tensor matrix, combined with the density-velocity response data of the fish escape behavior and the UAV dynamic constraints;

[0122] Solving module 703: for determining the individual trajectory basis functions of the UAVs using a set of bionic swarm interaction rules and a four-dimensional space-time gradient field tensor matrix, employing a space-time finite element discretization method combined with GPU-accelerated parallel solving;

[0123] Processing module 704: used to update data in real time based on individual drone trajectory basis functions and environmental sensors, dynamically adjust the four-dimensional space-time gradient field tensor matrix through Lyapunov function optimization processing, obtain the updated space-time gradient field, and generate group reorganization instructions;

[0124] Generation module 705: used to integrate the updated spatiotemporal gradient field, group reorganization instructions and historical trajectory data, apply the quantum annealing-NSGAIII hybrid optimization algorithm, and generate a bionic group trajectory emergence solution for drones enabled by the spatiotemporal gradient field.

[0125] Specifically, the acquisition module 701 includes:

[0126] Acquisition unit: used to acquire ice surface images through synthetic aperture radar carried by polar-orbiting satellites, extract crack distribution binary maps with a spatial resolution of ≤1m through ice crack edge detection algorithms, and obtain satellite remote sensing ice crack distribution data; collect raw point cloud data through solid-state lidar carried by drones, register multiple frame point clouds through point cloud registration algorithms, generate digital elevation models, and obtain LiDAR real-time terrain scanning data; measure horizontal flow velocity vectors in real time based on Doppler current meters built into terrain sensor arrays with a spacing of 500m, and generate ocean current velocity field data through Kriging spatial interpolation; obtain ice surface temperature distribution matrix at a sampling rate of 5Hz through a high-resolution infrared thermal imager carried by drones, and calculate spatial gradient fields based on the Sobel operator to obtain surface temperature gradient data; obtain instantaneous wind speed vector data by performing Kriging interpolation calculations on the three-dimensional wind speeds obtained from the complex environment meteorological station network;

[0127] The first construction unit is used to encode satellite remote sensing ice crack distribution data, LiDAR real-time terrain scanning data, ocean current velocity field data, surface temperature gradient data and instantaneous wind speed vector data into a five-channel space-time tensor, and use the five-channel space-time tensor as the input of the dynamic convolutional network to construct a dynamic convolutional neural network. The convolution kernel weight is jointly modulated by the ice crack curvature and temperature gradient, and the output is a coupled drift field vector. Combined with the coordinates and priority weights provided by the task requirement system, a four-dimensional space-time gradient field tensor matrix is ​​generated, where the four-dimensional space-time gradient field tensor matrix includes the gravitational field, the repulsive field and the drift field.

[0128] Specifically, the building block 702 includes:

[0129] The first generation unit is used to fit the nonlinear mapping relationship between density and speed based on the group density-speed observation data of the fish escape behavior using a Sigmoid function to generate a bionic density-speed response constraint curve. The calculation formula is as follows:

[0130]

[0131] Where, is the density-velocity response constraint curve, is the maximum allowed speed of the drone, is the drone population density, is the critical density threshold, k is the steepness coefficient of the Sigmoid curve, e is the base of the natural logarithm, ;

[0132] The second construction unit is used to construct the Hamiltonian function by combining the bionic density-velocity response constraint curve with the preset UAV dynamics model, and generate the dynamic feasible solution space of the nonlinear differential game. The UAV dynamics model includes the integration of polar parameter measurements, gradient field coupling, and bionic rule constraints. The calculation formula for constructing the Hamiltonian function is as follows:

[0133]

[0134] Where, For drone control input, is the Hamiltonian function, is the co-state vector, T is the matrix, is the state equation of the UAV, is the control input matrix;

[0135] The second generation unit is used to generate the swarm obstacle avoidance and collision avoidance joint force vector field based on the ice crack repulsion gradient field component in the four-dimensional space-time gradient field tensor matrix, integrating the real-time relative position of the drone group, and using the inverse proportional repulsion potential energy superposition algorithm;

[0136] Optimization unit: It is used to obtain the bionic swarm interaction rule set, including static constraints and dynamic collaborative strategies, based on the data of the dynamic feasible solution space, the swarm obstacle avoidance and collision avoidance combined force vector field, and the drift field, through the solution of the Bourn-Nash equilibrium and the collaborative optimization of the heading.

[0137] Specifically, the second generating unit includes:

[0138] The first calculation unit is used to calculate the ice crack repulsion gradient field covering the global task area based on the ice crack repulsion gradient field component in the four-dimensional space-time gradient field tensor matrix;

[0139] The second calculation unit is used to extract the local repulsive force gradient in the ice crack repulsive force gradient field, combine it with the real-time position of the UAV, and generate the anti-collision force between groups through inverse proportional potential energy superposition. The calculation formula is as follows:

[0140]

[0141] Where, 、 denote the positions of drones i and j respectively, γ =0.8 is the repulsive force weight coefficient, For drones i The inter-group collision avoidance force, Euclidean distance between two drones;

[0142] The third calculation unit is used to superimpose the local repulsive force gradient and the anti-collision force vector between the groups to obtain the total obstacle avoidance and collision avoidance combined force of the individual drones. The calculation formula is as follows:

[0143]

[0144] Where, is the local repulsive gradient, To prevent collisions between groups, is the combined force vector field of group obstacle avoidance and collision avoidance.

[0145] Specifically, the solution module 703 includes:

[0146] Constructor set unit: It is used to construct a spatiotemporal threat perception basis function set based on the four-dimensional spatiotemporal gradient field tensor matrix and the bionic swarm heading coordination constraint, using the Fourier-Legendre joint basis function expansion method. Its spatial wavenumber component is modulated by the ice crack distribution density, and the temporal frequency component is associated with the drift field time-varying rate. The heading coordination angle is embedded in the basis function phase through the cosine modulation term. Its calculation formula is as follows:

[0147]

[0148] Where, For the k spatiotemporal coupling basis functions, is the spatial fluctuation term, which characterizes the periodic change of the basis function in the horizontal plane. is the spatial wave number modulated by the ice crack distribution, is the time frequency parameter, is the group heading coordination angle, is the space-time gradient field tensor G The modulus length reflects the intensity of environmental threats. is the group heading coordination angle, t is the time variable;

[0149] Discrete generation unit: used to discretize the spatiotemporal threat perception basis function set and the UAV dynamics model through the Galerkin weighted residual method to generate a trajectory coupled algebraic system , where the coefficient matrix It is composed of the inner product of basis functions, and the right-hand term Contains orthogonal projections of group collision avoidance forces;

[0150]

[0151] Where, is the integral operation in the space-time domain, For drones i The partial derivative of the trajectory function with respect to time t, is the negative gradient of the ice crack repulsive field, is the environmental drift field, For the control input of the drone, is the kth space-time basis function, is the product of the infinitesimal elements of space and time;

[0152] Partitioning strategy unit: This unit implements GPU parallel processing for ice crevasse risk classification based on a trajectory coupled algebraic system and a four-dimensional space-time gradient field tensor matrix to obtain a risk zoning strategy. The risk zoning strategy includes allocating 32×32 thread blocks to a high-risk zone within 0.5 m of the crevasse edge, a 1-2 m area outside the crevasse, a 64×64 thread block to a medium-risk zone and a flat ice surface, and a 128×128 thread block to a safe zone. Double-precision floating-point operations and preconditioning acceleration are used in the high-risk zone, while single-precision mixed iterative calculations are used in the medium- and low-risk zones.

[0153] Output generation unit: used to output the basis function coefficient vector through the risk partitioning strategy to generate the UAV continuous trajectory function.

[0154] Specifically, the generating module 705 includes:

[0155] The third construction unit is used to construct a multi-objective problem, including minimizing energy consumption, minimizing the number of ice crevasse collisions, and optimizing mission synchronization, based on the updated spatiotemporal gradient field, swarm reorganization instructions, and historical trajectory data. The control input is constrained by the maximum thrust and bionic speed of the UAV, ice crevasse collision determination is triggered by the curvature field threshold, and mission synchronization is measured by the arrival time difference.

[0156] The third generation unit is used to generate the initial population using the quantum annealing mechanism. Its Hamiltonian is designed to couple the historical trajectory conflict statistics with the repulsive field gradient direction alignment. Each individual encodes the control input sequence and heading strategy of the drone. The quantum bit state reflects the decision variable. The coupling strength is determined by the swarm coordination risk quantification. The local field strength is correlated with the current gradient field direction consistency.

[0157] Conversion unit: It is used to optimize the population through a hybrid strategy of quantum tunneling mutation and dynamic reference point adjustment. The final output Pareto optimal solution set is converted into a continuous trajectory scheme enabled by the spatiotemporal gradient field. Its coefficients are directly mapped by the optimization variables. The heading instructions and reorganization marks are inherited from the decision coding of the optimal solution. Its calculation formula is as follows:

[0158]

[0159] Where, is the optimized UAV trajectory function, K is the number of basis functions, is the optimal basis function coefficient, is the horizontal position x and the mission time t, Space-time basis functions.

[0160] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated on here.

[0161] Example 3:

[0162] Corresponding to the above method embodiment, this embodiment also provides a drone bionic group trajectory emergence device with a spatiotemporal gradient field. The drone bionic group trajectory emergence device with a spatiotemporal gradient field described below and the drone bionic group trajectory emergence method with a spatiotemporal gradient field described above can be referenced to each other.

[0163] Figure 3 FIG. 8 is a block diagram of a drone bionic group trajectory emergence device 800 of a spatiotemporal gradient field according to an exemplary embodiment. Figure 3 As shown, the drone biomimetic swarm trajectory emergence device 800 of the spatiotemporal gradient field includes: a processor 801 and a memory 802. The drone biomimetic swarm trajectory emergence device 800 of the spatiotemporal gradient field also includes one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0164] The processor 801 is used to control the overall operation of the spatiotemporal gradient field drone biomimetic swarm trajectory emergence device 800 to complete all or part of the steps in the spatiotemporal gradient field drone biomimetic swarm trajectory emergence method. The memory 802 is used to store various types of data to support the operation of the spatiotemporal gradient field drone biomimetic swarm trajectory emergence device 800. Such data may include, for example, instructions for any application or method operating on the spatiotemporal gradient field drone biomimetic swarm trajectory emergence device 800, as well as application-related data such as contact information, sent and received messages, pictures, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, and the above-mentioned other interface modules can be a keyboard, a mouse or buttons, etc. These buttons can be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the drone bionic group trajectory emergence device 800 and other devices in the spatiotemporal gradient field. Wireless communication, such as Wi-Fi, Bluetooth, near field communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 805 may include: a Wi-Fi module, a Bluetooth module or an NFC module.

[0165] In an exemplary embodiment, the drone biomimetic swarm trajectory emergence device 800 with a spatiotemporal gradient field can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the above-mentioned drone biomimetic swarm trajectory emergence method with a spatiotemporal gradient field.

[0166] In another exemplary embodiment, a computer-readable storage medium including program instructions is further provided. When executed by a processor, the program instructions implement the steps of the above-described method for the emergence of bionic swarm trajectories of drones using a spatiotemporal gradient field. For example, the computer-readable storage medium may be the aforementioned memory 802 including the program instructions. The program instructions may be executed by the processor 801 of the device 800 for the emergence of bionic swarm trajectories of drones using a spatiotemporal gradient field to implement the above-described method for the emergence of bionic swarm trajectories of drones using a spatiotemporal gradient field.

[0167] Example 4:

[0168] Corresponding to the above method embodiment, a readable storage medium is also provided in this embodiment. The readable storage medium described below and the drone bionic group trajectory emergence method of a spatiotemporal gradient field described above can be referenced to each other.

[0169] A computer program is stored on the readable storage medium, and when the computer program is executed by the processor, the steps of the method for the emergence of bionic swarm trajectories of drones in the spatiotemporal gradient field of the above-mentioned method embodiment are implemented.

[0170] The readable storage medium may specifically be any readable storage medium that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0171] In summary, through the real-time fusion of satellite, LiDAR and buoy array data using the four-dimensional space-time gradient field tensor matrix (repulsive field, gravitational field, drift field), a dynamic weighted model of the repulsive gradient at the edge of the ice crack is constructed, achieving millisecond-level updates of environmental threats; the density-velocity response curve is incorporated into the Hamiltonian function to generate distributed rules that take into account both biological rationality and dynamic feasibility, reducing the communication load to 0.5Mbps, and using the Fourier-Legendre joint basis function expansion. Through risk-graded parallel computing, the time required to solve the trajectory of a thousand-aircraft group is compressed to within 5ms, and the accuracy in high-risk areas reaches 0.1m; the density of reference points in high-risk areas is increased by 3 times, combined with quantum tunneling variation, breaking through local optimality, and the Pareto solution set covering all dimensions of energy consumption, safety and synchronization. The mission success rate is increased to 98.7%, providing polar drone swarms with high-precision, low-latency and strong robust autonomous navigation capabilities, which is significantly better than the existing technology level.

[0172] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

[0173] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for the emergence of bionic swarm trajectories of drones in a spatiotemporal gradient field, characterized by: include: Obtain satellite remote sensing ice crack distribution data, LiDAR real-time terrain scanning data, ocean current velocity field data, surface temperature gradient data and instantaneous wind speed vector data, and generate a four-dimensional spatiotemporal gradient field tensor matrix through a dynamic field coupled convolutional network; Based on the four-dimensional space-time gradient field tensor matrix, combined with the density-velocity response data of fish escape behavior and the dynamic constraints of the UAV, a set of bionic swarm interaction rules is constructed through nonlinear differential game modeling. Using the bionic swarm interaction rule set and the four-dimensional space-time gradient field tensor matrix, the space-time finite element discretization method is adopted and combined with GPU accelerated parallel solution to determine the individual trajectory basis functions of the UAV; Based on the real-time update data of the drone's individual trajectory basis functions and environmental sensors, the four-dimensional space-time gradient field tensor matrix is ​​dynamically adjusted through Lyapunov function optimization processing to obtain the updated space-time gradient field and generate group reorganization instructions; By integrating the updated spatiotemporal gradient field, group reorganization instructions and historical trajectory data, and applying the quantum annealing-NSGAIII hybrid optimization algorithm, a bionic group trajectory emergence scheme for drones enabled by the spatiotemporal gradient field is generated.

2. The method for emerging trajectories of drone bionic groups based on spatiotemporal gradient fields according to claim 1 is characterized in that: The satellite remote sensing ice crack distribution data, LiDAR real-time terrain scanning data, ocean current velocity field data, surface temperature gradient data and instantaneous wind speed vector data are obtained, and a four-dimensional spatiotemporal gradient field tensor matrix is ​​generated through a dynamic field coupled convolutional network, which includes: Ice surface images are acquired using synthetic aperture radars carried by polar-orbiting satellites. A crack edge detection algorithm is used to extract binary crack distribution maps with a spatial resolution of ≤1m, resulting in satellite remote sensing ice crack distribution data. Raw point cloud data is collected using a solid-state lidar carried by an unmanned aerial vehicle (UAV). Multi-frame point clouds are registered using a point cloud registration algorithm to generate a digital elevation model, resulting in real-time LiDAR terrain scanning data. Doppler velocity meters built into a 500m-spaced terrain sensor array measure horizontal flow velocity vectors in real time, and ocean current velocity field data are generated using Kriging spatial interpolation. A high-resolution infrared thermal imager carried by the UAV acquires the ice surface temperature distribution matrix at a sampling rate of 5Hz, and the spatial gradient field is calculated using the Sobel operator to obtain surface temperature gradient data. Instantaneous wind speed vector data is obtained by performing Kriging interpolation on the three-dimensional wind speeds obtained from a network of complex weather stations. Satellite remote sensing ice crack distribution data, LiDAR real-time terrain scanning data, ocean current velocity field data, surface temperature gradient data and instantaneous wind speed vector data are encoded into a five-channel space-time tensor, and the five-channel space-time tensor is used as the input of the dynamic convolutional network to construct a dynamic convolutional neural network. Its convolution kernel weights are jointly modulated by the ice crack curvature and temperature gradient, and the output is a coupled drift field vector. Combined with the coordinates and priority weights provided by the mission requirement system, a four-dimensional space-time gradient field tensor matrix is ​​generated, where the four-dimensional space-time gradient field tensor matrix includes the gravitational field, repulsive field and drift field.

3. The method for emerging trajectory of drone bionic swarms based on spatiotemporal gradient field according to claim 1, characterized in that: Based on the four-dimensional space-time gradient field tensor matrix, combined with the density-velocity response data of fish escape behavior and the UAV dynamic constraints, a set of bionic swarm interaction rules is constructed through nonlinear differential game modeling, including: Based on the group density-velocity observation data of fish escape behavior, the Sigmoid function is used to fit the nonlinear mapping relationship between density and velocity to generate a bionic density-velocity response constraint curve. The calculation formula is as follows: Where, is the density-velocity response constraint curve, is the maximum allowed speed of the drone, is the drone population density, is the critical density threshold, k is the steepness coefficient of the Sigmoid curve, e is the base of the natural logarithm, ; The bionic density-velocity response constraint curve is combined with the preset UAV dynamics model to construct a Hamiltonian function, generating a dynamic feasible solution space for the nonlinear differential game. The UAV dynamics model is obtained by integrating polar parameter measurements, gradient field coupling, and bionic rule constraints. The calculation formula for constructing the Hamiltonian function is as follows: Where, For drone control input, is the Hamiltonian function, is the co-state vector, T is the matrix, is the state equation of the UAV, is the control input matrix; Based on the ice crack repulsion gradient field component in the four-dimensional space-time gradient field tensor matrix, the real-time relative position of the drone group is integrated, and the inverse proportional repulsion potential energy superposition algorithm is used to generate the group obstacle avoidance and collision avoidance joint force vector field; Based on the data of the dynamic feasible solution space, the combined force vector field of swarm obstacle avoidance and collision avoidance, and the drift field, a set of bionic swarm interaction rules is obtained through Bourns-Nash equilibrium solution and heading collaborative optimization, which includes static constraints and dynamic collaborative strategies.

4. The method for emerging trajectory of drone bionic swarms based on spatiotemporal gradient fields according to claim 3 is characterized in that: Based on the ice crack repulsion gradient field component in the four-dimensional space-time gradient field tensor matrix, the real-time relative position of the drone group is integrated, and the inverse proportional repulsion potential energy superposition algorithm is used to generate the group obstacle avoidance and collision avoidance joint force vector field, which includes: Based on the ice crack repulsion gradient field component in the four-dimensional space-time gradient field tensor matrix, the ice crack repulsion gradient field covering the global mission area is calculated; The local repulsive gradient in the ice crack repulsive gradient field is extracted and combined with the real-time position of the UAV to generate the anti-collision force between groups through inverse proportional potential energy superposition. The calculation formula is as follows: Where, 、 denote the positions of drones i and j respectively, γ =0.8 is the repulsive force weight coefficient, For drones i The inter-group collision avoidance force, Euclidean distance between two drones; The local repulsive force gradient and the anti-collision force between groups are superimposed to obtain the total obstacle avoidance and collision avoidance combined force of the individual drones. The calculation formula is as follows: Where, is the local repulsive gradient, To prevent collisions between groups, is the combined force vector field of group obstacle avoidance and collision avoidance.

5. The method for emerging bionic swarm trajectories of drones in spatiotemporal gradient fields according to claim 3 is characterized in that: The bionic swarm interaction rule set and the four-dimensional space-time gradient field tensor matrix are used to determine the individual trajectory basis functions of the drones by adopting the space-time finite element discretization method and combining it with GPU accelerated parallel solution, including: Based on the four-dimensional space-time gradient field tensor matrix and the bionic swarm heading coordination constraint, the Fourier-Legendre joint basis function expansion method is used to construct a space-time threat perception basis function set. Its spatial wavenumber component is modulated by the ice crack distribution density, and the temporal frequency component is associated with the drift field time-varying rate. The heading coordination angle is embedded in the basis function phase through the cosine modulation term. Its calculation formula is as follows: Where, For the k spatiotemporal coupling basis functions, is the spatial fluctuation term, which characterizes the periodic change of the basis function in the horizontal plane. is the spatial wave number modulated by the ice crack distribution, is the time frequency parameter, is the group heading coordination angle, is the space-time gradient field tensor G The modulus length reflects the intensity of environmental threats. is the group heading coordination angle, t is the time variable; The spatiotemporal threat perception basis function set and the UAV dynamics model are discretized using the Galerkin weighted residual method to generate a trajectory coupled algebraic system. , where the coefficient matrix It is composed of the inner product of basis functions, and the right-hand term The orthogonal projection of the group collision avoidance force is included, and the calculation formula of the discretization of the Galerkin weighted residual method is as follows: Where, is the integral operation in the space-time domain, For drones i The partial derivative of the trajectory function with respect to time t, is the negative gradient of the ice crack repulsive field, is the environmental drift field, For the control input of the drone, is the kth space-time basis function, is the product of the infinitesimal elements of space and time; Based on a trajectory-coupled algebraic system and a four-dimensional space-time gradient field tensor matrix, a GPU-parallel solution for ice crevasse risk classification was implemented to obtain a risk zoning strategy. The strategy includes allocating 32×32 thread blocks to a high-risk zone within 0.5 m of the crevasse edge, a 1-2 m area outside the crevasse, a 64×64 thread block to a medium-risk zone and a flat ice surface, and a 128×128 thread block to a safe zone. Double-precision floating-point operations and preconditioning acceleration are used in the high-risk zone, while single-precision mixed iterative calculations are used in the medium- and low-risk zones. The basis function coefficient vector is output through the risk partitioning strategy to generate the UAV continuous trajectory function.

6. The method for emerging trajectory of drone bionic swarms based on spatiotemporal gradient fields according to claim 1, characterized in that: The data is updated in real time based on the individual trajectory basis function of the drone and the environmental sensor, and is optimized by the Lyapunov function. The calculation formula of the Lyapunov function optimization is as follows: Where, L ( t ) is a scalar function that characterizes the overall trajectory tracking error of the UAV swarm, is the actual position of UAV i, is the predicted position of UAV i, is the actual speed of UAV i, is the predicted speed of UAV i, λ is the speed error weight coefficient, and N is the size of the UAV swarm.

7. The method for emerging trajectory of drone bionic swarms based on spatiotemporal gradient fields according to claim 1, characterized in that: The quantum annealing-NSGAIII hybrid optimization algorithm is applied to the integrated updated spatiotemporal gradient field, group reorganization instructions and historical trajectory data to generate a spatiotemporal gradient field-enabled UAV biomimetic group trajectory emergence scheme, which includes: Based on the updated spatiotemporal gradient field, swarm reorganization instructions, and historical trajectory data, a multi-objective problem was constructed, including minimizing energy consumption, minimizing the number of ice crevasse collisions, and optimizing mission synchronization. The control input was constrained by the maximum thrust and bionic speed constraints of the UAV, ice crevasse collision determination was triggered by the curvature field threshold, and mission synchronization was measured by the arrival time difference. The initial population is generated using a quantum annealing mechanism. Its Hamiltonian is designed to couple historical trajectory conflict statistics with the alignment of the repulsive field gradient direction. Each individual encodes the control input sequence and heading strategy of the drone. The quantum bit state reflects the decision variable. The coupling strength is determined by the swarm's collaborative risk quantification. The local field strength is correlated with the current gradient field direction consistency. The population is optimized by a hybrid strategy of quantum tunneling mutation and dynamic reference point adjustment. The final output Pareto optimal solution set is converted into a continuous trajectory scheme enabled by the spatiotemporal gradient field. Its coefficients are directly mapped by the optimization variables, and the heading instructions and reorganization marks are inherited from the decision coding of the optimal solution. Its calculation formula is as follows: Where, is the optimized UAV trajectory function, K is the number of basis functions, is the optimal basis function coefficient, is the horizontal position x and the mission time t, Space-time basis functions.

8. A UAV bionic swarm trajectory emergence system based on a spatiotemporal gradient field, based on the UAV bionic swarm trajectory emergence method based on a spatiotemporal gradient field according to claim 1, characterized in that: include: Acquisition module: used to obtain satellite remote sensing ice crack distribution data, LiDAR real-time terrain scanning data, ocean current velocity field data, surface temperature gradient data and instantaneous wind speed vector data, and generate a four-dimensional space-time gradient field tensor matrix through a dynamic field coupled convolutional network; Building module: Used to construct a set of biomimetic swarm interaction rules through nonlinear differential game modeling based on the four-dimensional space-time gradient field tensor matrix, combined with the density-velocity response data of fish escape behavior and the dynamic constraints of the drone; Solution module: Used to determine the individual trajectory basis functions of UAVs using the bionic swarm interaction rule set and the four-dimensional space-time gradient field tensor matrix, using the space-time finite element discretization method and GPU-accelerated parallel solution; Processing module: used to update data in real time based on the individual trajectory basis functions of drones and environmental sensors, dynamically adjust the four-dimensional space-time gradient field tensor matrix through Lyapunov function optimization processing, obtain the updated space-time gradient field and generate group reorganization instructions; Generation module: It is used to integrate the updated spatiotemporal gradient field, group reorganization instructions and historical trajectory data, and apply the quantum annealing-NSGAIII hybrid optimization algorithm to generate a bionic group trajectory emergence solution for drones enabled by the spatiotemporal gradient field.

9. The UAV bionic swarm trajectory emergence system of spatiotemporal gradient field according to claim 8, characterized in that: The acquisition module includes: Acquisition unit: used to acquire ice surface images through synthetic aperture radar carried by polar-orbiting satellites, extract crack distribution binary maps with a spatial resolution of ≤1m through ice crack edge detection algorithms, and obtain satellite remote sensing ice crack distribution data; collect raw point cloud data through solid-state lidar carried by drones, register multiple frame point clouds through point cloud registration algorithms, generate digital elevation models, and obtain LiDAR real-time terrain scanning data; measure horizontal flow velocity vectors in real time based on Doppler current meters built into terrain sensor arrays with a spacing of 500m, and generate ocean current velocity field data through Kriging spatial interpolation; obtain ice surface temperature distribution matrix at a sampling rate of 5Hz through a high-resolution infrared thermal imager carried by drones, and calculate spatial gradient fields based on the Sobel operator to obtain surface temperature gradient data; obtain instantaneous wind speed vector data by performing Kriging interpolation calculations on the three-dimensional wind speeds obtained from the complex environment meteorological station network; The first construction unit is used to encode satellite remote sensing ice crack distribution data, LiDAR real-time terrain scanning data, ocean current velocity field data, surface temperature gradient data and instantaneous wind speed vector data into a five-channel space-time tensor, and use the five-channel space-time tensor as the input of the dynamic convolutional network to construct a dynamic convolutional neural network. The convolution kernel weight is jointly modulated by the ice crack curvature and temperature gradient, and the output is a coupled drift field vector. Combined with the coordinates and priority weights provided by the task requirement system, a four-dimensional space-time gradient field tensor matrix is ​​generated, where the four-dimensional space-time gradient field tensor matrix includes the gravitational field, the repulsive field and the drift field.

10. The UAV bionic swarm trajectory emergence system of spatiotemporal gradient field according to claim 8, characterized in that: The building blocks include: The first generation unit is used to fit the nonlinear mapping relationship between density and speed based on the group density-speed observation data of the fish escape behavior using a Sigmoid function to generate a bionic density-speed response constraint curve. The calculation formula is as follows: Where, is the density-velocity response constraint curve, is the maximum allowed speed of the drone, is the drone population density, is the critical density threshold, k is the steepness coefficient of the Sigmoid curve, e is the base of the natural logarithm, ; The second construction unit is used to construct the Hamiltonian function by combining the bionic density-velocity response constraint curve with the preset UAV dynamics model, and generate the dynamic feasible solution space of the nonlinear differential game. The UAV dynamics model includes the integration of polar parameter measurements, gradient field coupling, and bionic rule constraints. The calculation formula for constructing the Hamiltonian function is as follows: Where, For drone control input, is the Hamiltonian function, is the co-state vector, T is the matrix, is the state equation of the UAV, is the control input matrix; The second generation unit is used to generate the swarm obstacle avoidance and collision avoidance joint force vector field based on the ice crack repulsion gradient field component in the four-dimensional space-time gradient field tensor matrix, integrating the real-time relative position of the drone group, and using the inverse proportional repulsion potential energy superposition algorithm; Optimization unit: It is used to obtain the bionic swarm interaction rule set, including static constraints and dynamic collaborative strategies, based on the data of the dynamic feasible solution space, the swarm obstacle avoidance and collision avoidance combined force vector field, and the drift field, through the solution of the Bourn-Nash equilibrium and the collaborative optimization of the heading.

Images