Low-altitude unmanned aerial vehicle traffic flow equalization distribution method and system adopting risk evolution

Abstract

The application discloses a low-altitude unmanned aerial vehicle traffic flow equalization distribution method and system using risk evolution, and relates to the technical field of low-altitude traffic management. The method comprises the following steps: acquiring airspace traffic flow topology, path design capacity and airspace inherent risk field; performing isolated path decision for each unmanned aerial vehicle based on the inherent risk field, and forming a virtual path-probability matrix; performing single-time risk pressure evolution evaluation based on the virtual path-probability matrix, quantifying the dynamic influence of traffic flow aggregation on the risk field, and configuring a discard probability for each path; and finally, fusing the airspace inherent risk field, the virtual path-probability matrix and the discard probability, performing iterative optimization solution by using an optimization algorithm, and outputting a target traffic flow distribution scheme with the minimum total cost. Through the interactive feedback of the dynamic risk evolution mechanism and the group path selection behavior, the overall optimization of the traffic flow distribution is realized, and the safety and efficiency of the collaborative operation of the low-altitude unmanned aerial vehicles are improved.

Description

Technical Field

[0001] This invention relates to the field of low-altitude traffic management technology, specifically to a method and system for balancing low-altitude unmanned aerial vehicle (UAV) traffic flow using risk evolution. Background Technology

[0002] The application of low-altitude drones continues to expand, and the airspace operating environment is becoming increasingly complex. Low-altitude airspace contains both static obstacles such as buildings and high-voltage cables, and is also affected by dynamic factors such as weather conditions and temporary no-fly zones, creating a complex inherent risk field. When multiple drones operate collaboratively, their path selection behaviors are interconnected, and the cumulative effect of individual decisions can easily lead to overload of traffic on specific routes, causing local congestion and safety risks.

[0003] Existing technologies mostly focus on single-drone path planning, using static environment models to calculate the optimal path for an independent UAV. These methods fail to fully consider the dynamic interaction characteristics of traffic flow in multi-drone cooperative scenarios and cannot reflect the feedback impact of group selection behavior on the airspace risk field. When a large number of UAVs choose paths based on similar strategies, traffic flow aggregation occurs, increasing the probability of collisions and amplifying existing risks, leading to a decrease in overall system efficiency. Traditional static allocation models are ill-suited to this dynamic evolution process and cannot achieve an effective balance between safety and efficiency. Summary of the Invention

[0004] This invention addresses the technical problem in existing technologies that cannot effectively handle the dynamic aggregation of traffic flow and the resulting amplification of airspace risks during the collaborative operation of multiple UAVs, making it difficult to achieve a balance between safety and efficiency. It provides a low-altitude UAV traffic flow equalization allocation method and system that adopts risk evolution.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0006] In a first aspect, the present invention provides a low-altitude unmanned aerial vehicle (UAV) traffic flow equilibrium allocation method employing risk evolution, comprising:

[0007] Obtain the airspace traffic flow topology of the target area, and determine the path design capacity and inherent airspace risk field of each traffic flow path in the airspace traffic flow topology.

[0008] Based on the inherent risk field of the airspace, an isolated path decision is made for each UAV to form a virtual path-probability matrix;

[0009] A one-time risk pressure evolution assessment is performed based on the virtual path-probability matrix, and an abandonment probability is configured for each traffic flow path according to the risk pressure evolution assessment results.

[0010] Using the inherent risk field of the airspace, the virtual path-probability matrix, and the rejection probability, an optimization algorithm is used to iteratively optimize the traffic flow allocation and output the converged result as the target traffic flow allocation scheme.

[0011] Secondly, the present invention provides a low-altitude unmanned aerial vehicle (UAV) traffic flow equalization allocation system employing risk evolution, comprising:

[0012] The airspace topology and risk field acquisition module is used to acquire the airspace traffic flow topology of the target area and determine the path design capacity and inherent airspace risk field of each traffic flow path in the airspace traffic flow topology.

[0013] An isolated path decision module is used to make isolated path decisions for each UAV based on the inherent risk field of the airspace, forming a virtual path-probability matrix.

[0014] The risk pressure assessment and abandonment probability configuration module is used to perform a one-time risk pressure evolution assessment based on the virtual path-probability matrix, and configure the abandonment probability for each traffic flow path according to the risk pressure evolution assessment results.

[0015] The balanced allocation optimization solution module is used to iteratively optimize the balanced allocation of traffic flow using the inherent risk field of the airspace, the virtual path-probability matrix and the rejection probability, and output the converged result as the target traffic flow allocation scheme.

[0016] The beneficial effects of this invention are:

[0017] Compared to existing technologies, this invention first establishes a precise quantitative foundation for the static safety carrying capacity of airspace by introducing the inherent risk field of airspace and path design capacity. Secondly, it uses isolated path decision-making to generate a virtual path-probability matrix, effectively simulating the initial path selection tendency of UAV swarms in a risky environment, providing data support for subsequent risk evolution analysis. Thirdly, through a single-step risk pressure evolution assessment, it dynamically reveals the amplification effect of traffic flow aggregation on the airspace risk field, and generates abandonment probabilities accordingly, realizing a linkage feedback mechanism between risk and traffic pressure. Finally, it comprehensively utilizes optimization algorithms for iterative solutions, ensuring that the final allocation scheme not only satisfies the user equilibrium principle but also significantly reduces the overall risk level, achieving synergistic optimization of safety and efficiency. Attached Figure Description

[0018] Figure 1 A schematic diagram of the process for the low-altitude unmanned aerial vehicle traffic flow equilibrium allocation method using risk evolution provided by the present invention;

[0019] Figure 2 A schematic diagram of the structure of the low-altitude unmanned aerial vehicle traffic flow equalization allocation system using risk evolution provided by the present invention.

[0020] In the attached diagram, the components represented by each number are as follows:

[0021] Module 11 for obtaining airspace topology and risk field, Module 12 for isolated path decision-making, Module 13 for risk pressure assessment and abandonment probability configuration, and Module 14 for balanced allocation optimization solution. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0024] In the description of this invention, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this invention is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed herein.

[0025] Example 1, as Figure 1 As shown, this embodiment of the invention provides a low-altitude unmanned aerial vehicle (UAV) traffic flow equilibrium allocation method employing risk evolution, including:

[0026] S10: Obtain the airspace traffic flow topology of the target area, and determine the path design capacity and airspace inherent risk field of each traffic flow path in the airspace traffic flow topology.

[0027] Specifically, the airspace traffic flow topology of the target area is obtained, and the path design capacity and inherent airspace risk field of each traffic flow path in the airspace traffic flow topology are determined, including:

[0028] Based on the low-altitude traffic planning information of the target area, the airspace traffic flow topology is obtained, and the path design information corresponding to each traffic flow path in the airspace traffic flow topology is matched to extract the path design capacity, path level features and path speed limit features.

[0029] Based on the path level characteristics and the path speed limit characteristics, a preset path risk management plan is queried to obtain the risk management radius corresponding to each traffic flow path.

[0030] Using the risk management radius as the risk field boundary, the neighborhood risk factor set of the airspace traffic flow topology is collected by traversing the area, and the inherent risk field of the airspace is calculated based on the neighborhood risk factor set and the path risk management plan.

[0031] Specifically, the target area is a specific three-dimensional spatial range for carrying out low-altitude drone flight activities, such as urban logistics and distribution airspace, industrial inspection routes, or emergency rescue channels; airspace traffic flow topology refers to the networked path structure composed of multiple nodes and edges connecting the nodes, used to describe the spatial connection relationship of all legal flight paths in the airspace, obtained by parsing official airspace planning maps or digital geographic information, and used to characterize all physical access options available to the drone swarm.

[0032] Based on the airspace traffic flow topology, the path design capacity and inherent airspace risk field of each traffic flow path can be determined. Specifically, the path design capacity of each traffic flow path refers to the maximum UAV traffic threshold that the path can handle per unit time, provided that safety separation and control rules are met. This is usually pre-set by the airspace management authority based on the path's physical dimensions, communication and navigation performance, and operational rules. The inherent airspace risk field refers to the basic risk distribution field assessed based on the static environmental characteristics surrounding the path. It reflects the degree of impact of relatively fixed risk sources such as terrain obstacles, building distribution, electromagnetic environment, and permanent no-fly zones on path safety. These factors are combined to provide objective and quantitative safety benchmarks and capacity constraints for subsequent traffic flow allocation decisions.

[0033] First, based on the low-altitude traffic planning information of the target area, the airspace traffic flow topology of the target area is obtained. Specifically, this requires extracting the network of flight paths that have been officially planned and certified from the low-altitude traffic management database. After obtaining the airspace traffic flow topology, each traffic flow path needs to be precisely matched with the detailed path design information stored in the background database. Once the match is successful, three core parameters can be extracted: path design capacity, path level characteristics, and path speed limit characteristics. Among them, the path design capacity represents the maximum UAV traffic that the path can safely accommodate per unit time; the path level characteristics reflect the importance and functional classification of the path in the airspace network; and the path speed limit characteristics specify the maximum permissible speed for flight on the path.

[0034] Secondly, based on the extracted path level characteristics and path speed limit characteristics, the pre-defined path risk management plan is queried. The path risk management plan is a pre-established rule base that clearly defines the safety control standards corresponding to paths with different levels and speed limits. By querying this path risk management plan, a risk management radius can be set for each traffic flow path. This risk management radius defines the spatial boundary extending to both sides from the path centerline as the reference point, representing the area requiring risk monitoring and management; it serves as the geographical basis for subsequent risk field calculations.

[0035] Furthermore, using the defined risk management radius as the spatial boundary, the entire airspace traffic flow topology is traversed. During the traversal, all potential risk factors within the boundary defined by the risk management radius of each traffic flow path are collected, collectively forming a neighborhood risk factor set. This neighborhood risk factor set is a comprehensive dataset, containing information on various static and quasi-static risk sources, including building height and density around the path, distribution of geographical obstacles, electromagnetic environment intensity, meteorologically sensitive areas, and population activity density.

[0036] Based on the collected neighborhood risk factor set, and combined with the risk assessment rules and algorithm weights specified in the path risk management plan, the inherent risk field of the airspace is calculated. This calculation process first standardizes the neighborhood risk factor set to eliminate the influence of different dimensions. Then, based on the weight coefficients preset for each type of risk factor in the path risk management plan, a weighted fusion calculation is performed. For each traffic flow path, the calculation of its inherent risk value considers not only the intensity of the risk factor itself but also the Euclidean distance between the risk factor and each traffic flow path. A distance decay function can be used to characterize the characteristic that risk decreases with increasing distance. Finally, a spatial interpolation algorithm expands the risk values ​​on discrete paths into a continuous risk intensity distribution field covering the entire target airspace.

[0037] The final calculated inherent risk field of the airspace is a spatial data field that covers the entire target airspace and can quantitatively characterize its basic safety status. It stores the risk intensity value corresponding to each geographic coordinate point in a gridded form, which can provide core safety benchmark input for subsequent traffic flow allocation decisions.

[0038] S20: Based on the inherent risk field of the airspace, make isolated path decisions for each UAV to form a virtual path-probability matrix;

[0039] Based on the inherent risk field of the airspace, an isolated path decision is made for each UAV, forming a virtual path-probability matrix, including:

[0040] Define the path selection abundance n of the UAV, where n is less than 1;

[0041] Iterate through multiple drones, make isolated path decisions based on the inherent risk field of the airspace, and output each isolated path decision as a candidate path group. Obtain multiple candidate path groups corresponding to multiple drones, wherein each candidate path group includes multiple candidate paths, and each candidate path is associated with a candidate probability.

[0042] Using the path selection abundance n as a probability constraint, traverse multiple candidate path groups, serialize multiple candidate paths in descending order based on the candidate probabilities, perform probability cumulative analysis accordingly, select the top few candidate paths whose probability cumulative analysis results satisfy the probability constraints, and output them as virtual paths.

[0043] The virtual path-probability matrix is ​​constructed by integrating multiple virtual paths with their corresponding alternative probabilities.

[0044] First, a path selection abundance parameter *n* needs to be defined. This parameter is a value less than 1 and is used to control the breadth of decision-making scope for a single UAV when making path selections. The setting of path selection abundance *n* reflects the constraint on the tolerance for path diversity during the decision-making process. It is set according to the balance requirements of computational efficiency and decision coverage in actual application scenarios. For example, setting it to 0.8 indicates coverage of the main probability distribution to simplify calculations, while setting it to 0.95 aims for a more comprehensive assessment of decision probability.

[0045] Secondly, multiple drones are traversed, and isolated path decisions are made independently for each drone. Specifically, the decision-making process takes the inherent risk field of the airspace as input. Under isolated conditions, without considering the choices of other drones, multiple feasible paths from the starting point to the destination are calculated. Each isolated path decision outputs a set of alternative paths, which contains multiple alternative paths, and a corresponding alternative probability is calculated for each alternative path. This alternative probability characterizes the likelihood of the drone choosing that path when only the inherent risk field is considered.

[0046] After completing isolated decision-making for all UAVs, multiple candidate path groups corresponding to the number of UAVs are obtained. Then, using the path selection abundance *n* as a probability constraint, each candidate path group is processed sequentially. For a single UAV's candidate path group, the multiple candidate paths it contains are sorted in descending order according to their associated candidate probability values. The probability accumulation calculation is performed on the sorted path sequence, and the top few candidate paths whose probability accumulation analysis results satisfy the probability constraint (i.e., exceeding the path selection abundance *n*) are selected as virtual paths. In summary, by simplifying the initial candidate set through probability significance screening, it is ensured that the selected virtual paths cover the main possibilities of the decision. This screening mechanism effectively eliminates marginal path options with extremely low probability of occurrence while retaining high-probability core paths, thus significantly improving subsequent computational efficiency while maintaining decision accuracy.

[0047] Finally, the virtual path selection information of all drones is integrated. The final set of virtual paths determined by each drone and its corresponding alternative probabilities are combined to construct a virtual path-probability matrix. This matrix is ​​structured with drones as rows and virtual paths as columns, and the matrix elements represent the probability of a specific drone choosing a specific virtual path. This comprehensively characterizes the path selection tendency of a drone swarm under isolated decision-making mode, providing a data foundation for subsequent risk evolution analysis.

[0048] S30: Perform a one-time risk pressure evolution assessment based on the virtual path-probability matrix, and configure the abandonment probability for each traffic flow path according to the risk pressure evolution assessment results;

[0049] First, a single-step risk pressure evolution assessment is performed based on the virtual path-probability matrix, including:

[0050] Based on the virtual path-probability matrix, the virtual path selection probabilities of all drones are aggregated, and the expected mechanical flow on each traffic flow path is calculated.

[0051] Based on the low-altitude traffic knowledge graph, a dynamic risk factor set is defined, and dynamic risk factor sets of multiple UAVs are collected according to the dynamic risk factor set.

[0052] Based on standard real-world risk data, the standard risk factor set and mechanical standard flow of the traffic flow path are extracted by combining the dynamic risk factor set, and a risk evolution model is constructed accordingly.

[0053] Input the expected mechanical flow and the set of dynamic risk factors into the risk evolution model to obtain the evolved risk field, and output the expected mechanical flow and the evolved risk field as the risk pressure evolution assessment result.

[0054] Specifically, a single-stage risk and pressure evolution assessment is conducted based on a virtual path-probability matrix. This assessment refers to the comprehensive simulation and quantitative analysis of the dynamic changes in airspace risk caused by the path selection behavior of a drone swarm at a specific time point. By aggregating swarm selection probabilities to predict traffic distribution and coupling dynamic environmental factors with the inherent risk field to simulate the risk transmission mechanism, it is possible to proactively identify critical risk areas and pressure bottlenecks that may form due to traffic flow aggregation. This provides a quantitative basis for the precise configuration of subsequent path abandonment probabilities, thereby achieving a decision-making upgrade from passively avoiding static risks to actively controlling dynamic risks.

[0055] First, based on the virtual path-probability matrix, the selection probabilities of all drones for each virtual path are aggregated. By summing the corresponding probabilities of all drones selecting the same traffic flow path, the expected mechanical flow of each traffic flow path under the current virtual selection situation is calculated. This expected mechanical flow refers to the drone traffic load that the path may bear without intervention.

[0056] Secondly, based on the low-altitude traffic knowledge graph, a dynamic risk factor set affecting airspace safety is defined. This dynamic risk factor set refers to abstract categories characterizing potential threats in the airspace environment that change over time. It is structured through entity relationships within the low-altitude traffic knowledge graph and includes real-time changing risk sources such as changes in weather conditions, temporary airspace control, and sudden electromagnetic interference. Based on this dynamic risk factor set, dynamic risk factors related to multiple UAV flight missions at the current moment are collected, such as real-time wind speed and direction data, coordinates of temporary no-fly zones, and communication signal strength indicators. This results in a dynamic risk factor set that can quantify the real-time environmental risk level.

[0057] Furthermore, based on standard real-world risk data and combined with the obtained dynamic risk factor set, baseline parameters for each traffic flow path under standard operating conditions are extracted, including the standard risk factor set and mechanical standard flow rate. The standard real-world risk data refers to a statistically representative airspace operation safety dataset accumulated through long-term observation under historical normal operating conditions, obtained through data cleaning and statistical analysis. The standard risk factor set refers to the set of typical risk factor values ​​corresponding to each path under baseline operating conditions; the mechanical standard flow rate refers to the median safe flow rate that a path can stably maintain under conditions without abnormal interference. Further, using the above baseline parameters, a risk evolution model is constructed. The risk evolution model is a mathematical framework characterizing the dynamic relationship between airspace risk intensity and traffic flow and environmental factors, used to describe the nonlinear evolution of the airspace risk field under the combined effects of flow changes and dynamic risk factors.

[0058] Specifically, the risk evolution model is constructed based on a baseline state established using standard real-world risk data, achieved by establishing a quantitative mapping relationship between traffic flow parameters, environmental risk factors, and the airspace risk field. The model's construction includes the following two core implementation paths:

[0059] The first implementation method employs a multi-factor mathematical function framework. This framework characterizes the risk evolution mechanism through parameterized mathematical expressions. Specifically, it establishes a functional relationship with expected machinery traffic and UAV dynamic indicators as input variables and collision risk intensity as the output value. The parameters to be determined in the function are obtained by fitting historical operational data, for example, by using nonlinear regression methods to perform parameter inversion on a standard real-world risk dataset. The multi-factor mathematical function framework is characterized by its clear structure and strong interpretability.

[0060] The second implementation method employs a machine learning model architecture. This architecture automatically learns the evolution patterns of risk through a data-driven approach. Specifically, it uses traffic characteristics, drone status characteristics, and environmental parameters from historical data as training inputs, and actual observed risk indicators as training targets, employing supervised learning algorithms for model training. The trained machine learning model architecture can predict safety indicators such as collision risks from new operational data.

[0061] It should be noted that the two implementation paths described above are technically equivalent and can both achieve the core functions of the risk evolution model. In practical applications, the appropriate solution can be selected based on the data completeness and accuracy requirements of the specific scenario.

[0062] For example, a risk evolution model is constructed using machine learning methods. Since there are complex nonlinear interactions between the risk field and multiple factors during the dynamic evolution of low-altitude traffic flow, and deep neural networks have significant advantages in capturing deep nonlinear relationships between high-dimensional features, a deep neural network model is chosen to construct this risk evolution model.

[0063] Specifically, this risk evolution model employs a feedforward neural network architecture, primarily consisting of an input layer, hidden layers, and an output layer. The input layer receives a preprocessed multi-dimensional feature vector, integrating key indicators of expected mechanical flow and a dynamic risk factor set. The hidden layers utilize a three-layer fully connected structure, with the number of neurons in each layer set to 128, 64, and 32 respectively. ReLU activation functions are used between layers to enhance the model's non-linear expressive power. To prevent overfitting, a Dropout layer is embedded after each hidden layer, with a dropout rate set between 0.2 and 0.4. The output layer uses a linear activation function to output the numerical value of the evolved risk field, representing the risk intensity.

[0064] During training, key hyperparameters included an initial learning rate of 0.001, 500 training epochs, and a batch size of 64. The learning rate was set based on stability considerations for the adaptive moment estimation algorithm; the number of training epochs ensured the model fully learned complex patterns in the data; and the batch size balanced training efficiency with gradient stability. Specifically, a supervised learning approach was adopted. Standard risk factor sets, standard mechanical flow rates, and corresponding actual observed risk field data for each time period were extracted from historical spatial operation data as training samples. The sample set was divided into training, validation, and test sets in a predetermined ratio of 7:2:1.

[0065] Furthermore, the feature vectors in the training set are used as network input, and the corresponding actual observed risk field data are used as supervision signals. The network weights are iteratively optimized through backpropagation. The mean squared error loss function is used to quantify the deviation between the predicted risk field and the actual risk field, and the training process is monitored through a validation set. When the value of the validation set loss function no longer decreases after several consecutive training epochs and the model prediction accuracy reaches a predetermined threshold, such as 90%, training is terminated, and a converged risk evolution model is obtained. This risk evolution model can accurately characterize the evolution law of the risk field under the combined effect of traffic pressure and dynamic risk factors, providing a reliable dynamic risk assessment basis for balanced traffic flow allocation.

[0066] Finally, the calculated expected traffic flow and the collected dynamic risk factor set are input into the constructed risk evolution model to obtain the evolved risk field under the influence of current traffic flow expectations and real-time risk factors. This evolved risk field characterizes the airspace risk distribution after a single risk pressure evolution assessment. Simultaneously, the expected traffic flow and the evolved risk field are output as the results of this risk pressure evolution assessment, providing a basis for subsequent allocation rejection probabilities.

[0067] Furthermore, based on the risk pressure evolution assessment results, an abandonment probability is configured for each of the aforementioned traffic flow paths, including:

[0068] Calculate the risk enhancement coefficient for each traffic flow path, whereby the risk enhancement coefficient is the ratio of the sum of the inherent risk field and the evolved risk field of the airspace to the inherent risk field of the airspace.

[0069] Calculate the flow pressure coefficient for each traffic flow path, where the flow pressure coefficient is the ratio of the expected mechanical flow to the path design capacity;

[0070] The risk enhancement coefficient and the flow pressure coefficient are input into a pre-constructed rejection probability function to calculate the rejection probability of each traffic flow path, wherein the rejection probability function is a monotonically increasing function with respect to the risk enhancement coefficient and the flow pressure coefficient.

[0071] First, the risk amplification coefficient for each traffic flow path is calculated. This coefficient is obtained by adding the inherent airspace risk field to the evolved risk field and then comparing it to the inherent airspace risk field. Specifically, the risk amplification coefficient = (inherent airspace risk field + evolved risk field) / inherent airspace risk field. This coefficient is a dimensionless value, and its physical meaning is to quantitatively represent the increase in risk of the path at the current moment compared to its static baseline risk level due to dynamic traffic flow aggregation and environmental influences. When the evolved risk field increases significantly, the risk amplification coefficient will increase accordingly, indicating that the safety status of the path is deteriorating.

[0072] Secondly, the flow pressure coefficient for each traffic flow path is calculated. This flow pressure coefficient is obtained by calculating the ratio of the expected mechanical flow to the path's design capacity. Specifically, the flow pressure coefficient = expected mechanical flow / path design capacity. This flow pressure coefficient is also a dimensionless value, and its function is to objectively measure the current traffic load faced by the path. When the expected mechanical flow approaches or exceeds the path's design capacity, the flow pressure coefficient increases, indicating that the path may experience congestion or a decrease in operational efficiency.

[0073] Finally, the calculated risk enhancement coefficient and flow pressure coefficient are used as input parameters and imported into a pre-constructed abandonment probability function for joint calculation, outputting the abandonment probability value of the corresponding traffic flow path. This abandonment probability function is a monotonically increasing function of the risk enhancement coefficient and the flow pressure coefficient. Its mathematical properties ensure that when either coefficient increases or both coefficients increase simultaneously, the function output value, i.e., the abandonment probability, shows a monotonically increasing trend. This function characteristic has a clear decision-making guidance role: when the overall risk level of a certain path increases significantly due to traffic aggregation or environmental changes, the priority of selecting that path can be actively reduced by increasing the abandonment probability, thereby inhibiting the excessive concentration of traffic flow on high-risk, high-load paths, prompting the drone swarm to spontaneously choose safer and smoother alternative paths, and achieving autonomous equilibrium of airspace traffic flow under risk constraints and optimization of overall operational efficiency.

[0074] Specifically, the abandonment probability output by the abandonment probability function refers to the quantitative recommendation value for drones to abandon a given traffic flow path under the current risk enhancement coefficient and traffic pressure coefficient. This abandonment probability value is between 0 and 1, and its magnitude directly reflects the unsuitability of the path in a dynamically evolving environment, transforming abstract risk assessment results into actionable path selection instructions. When the abandonment probability approaches 1, it indicates that the path is extremely unsuitable for flight due to significantly increased risk or severe traffic overload, and avoidance is strongly recommended; when the abandonment probability approaches 0, it indicates that the path is safe and unobstructed, and can be considered a priority choice.

[0075] S40: Using the inherent risk field of the airspace, the virtual path-probability matrix, and the rejection probability, an optimization algorithm is used to iteratively optimize the traffic flow allocation and output the converged result as the target traffic flow allocation scheme.

[0076] Using the inherent risk field of the airspace, the virtual path-probability matrix, and the rejection probability, an optimization algorithm is employed to iteratively optimize traffic flow allocation, outputting a converged result as the target traffic flow allocation scheme, including:

[0077] By combining the virtual path-probability matrix with the rejection probability, traffic flow paths are assigned to each UAV in sequence to obtain an initial allocation scheme;

[0078] Based on the inherent risk field of the airspace, the total allocation cost of the initial allocation scheme is calculated and obtained, wherein the total allocation cost includes the total risk cost and the total cost.

[0079] The initial allocation scheme is cross-mutated based on the discard probability, and the total allocation cost set is obtained by iterative calculation based on the cross-mutation result.

[0080] Based on the total allocation cost set, the allocation scheme corresponding to the one with the smallest total allocation cost is selected as the target traffic flow allocation scheme.

[0081] By employing an optimization algorithm based on the inherent airspace risk field, virtual path-probability matrix, and rejection probability, an iterative optimization solution for balanced traffic flow allocation is developed. Specifically, with the goal of minimizing total cost, potential allocation schemes are continuously generated, evaluated, and screened through simulations of natural optimization mechanisms such as biological evolution or swarm intelligence. The converged output is used as the target traffic flow allocation scheme, which is a set of UAV path allocations that achieves the optimal overall risk level and operating cost while satisfying airspace capacity and safety constraints. The specific steps are as follows:

[0082] First, by combining the virtual path-probability matrix and the rejection probability, an initial path selection is generated for each drone, forming an initial allocation scheme. Through probability sampling and rejection judgment mechanisms, it is ensured that the scheme respects both individual choice preferences and the global path state.

[0083] Specifically, by combining the virtual path-probability matrix and the rejection probability, traffic flow paths are assigned to each UAV in sequence to obtain an initial allocation scheme, including:

[0084] Obtain the virtual path and alternative probabilities of the target UAV from the virtual path-probability matrix;

[0085] Based on the aforementioned candidate probabilities, one of the multiple virtual paths of the target UAV is randomly sampled as the first candidate path;

[0086] Generate a random number within a predetermined range and compare the random number with the rejection probability corresponding to the first candidate path;

[0087] When the random number is greater than the rejection probability, the first candidate path is stored as the traffic flow path of the target UAV in the initial allocation scheme.

[0088] When the random number is less than or equal to the rejection probability, the first candidate path is abandoned and random sampling is performed again;

[0089] Repeat the operation until traffic flow path assignment is completed for each drone, and obtain the initial assignment scheme.

[0090] First, for the target UAV to be assigned, obtain the set of all possible virtual paths for the UAV and the corresponding candidate probability for each path from its corresponding virtual path-probability matrix entries. The candidate probability reflects the initial tendency of the UAV to choose each path in an isolated decision-making environment.

[0091] Secondly, a first-level random decision is made. Based on the obtained candidate probability distribution, a path is randomly selected from the set of virtual paths for the target drone as the first candidate path. This sampling increases the chances of selecting paths with high candidate probabilities.

[0092] Next, a second-level random decision is made: the rejection decision. A random number is generated that is uniformly distributed within a predetermined interval, such as the interval 0-1, to simulate a random event following a uniform distribution. This random number is compared with the rejection probability corresponding to the first candidate path itself: if the generated random number is greater than the rejection probability, it means that the path has passed the screening and is officially determined as the allocation path for the target UAV, and is stored in the initial allocation scheme; if the random number is less than or equal to the rejection probability, the first candidate path is abandoned, and the first-level random sampling step is repeated to select a new candidate path for the same decision.

[0093] This allocation process is executed sequentially for each UAV within the airspace. Through repeated iterations of the aforementioned two-level stochastic decision-making mechanism—namely, continuous path sampling based on individual preferences and discard judgments based on the global state—a definite traffic flow path is successfully assigned to each UAV in the entire swarm. Finally, the results of all individual path allocations are summarized to form an initial traffic flow allocation scheme. This method, through a probabilistic screening mechanism, effectively ensures that the generated initial traffic flow allocation scheme not only respects the individual path selection tendencies of each UAV based on its inherent risk field but also introduces proactive avoidance strategies for global evolution risks and potential traffic congestion from the very beginning of scheme construction, laying a good foundation for subsequent iterative optimization.

[0094] Furthermore, a quantitative evaluation standard for the merits of the proposed solutions is established. Combining the inherent risk field of the airspace, the total allocation cost of the initial allocation scheme is calculated. This total allocation cost is a comprehensive objective function, encompassing both total risk cost and total cost. The total risk cost quantifies the overall safety threat faced by all UAVs under this scheme due to traversing paths with different risk levels; its calculation relies on the inherent risk field of the airspace, which characterizes the basic risks of the paths. The total cost reflects the operational efficiency of the scheme, such as total flight distance or total energy consumption. The total allocation cost provides a clear mathematical objective for subsequent optimization.

[0095] Secondly, an iterative improvement mechanism is initiated. A rejection probability is introduced, and crossover and mutation operations are performed on the initial allocation scheme. Crossover operations reorganize the local structure of feasible solutions by exchanging path assignments between different UAVs; mutation operations randomly adjust the path selection of individual UAVs based on a certain probability, introducing new exploration directions. Crossover and mutation aim to expand the search space through a controllable perturbation strategy. Each crossover or mutation operation generates a new candidate allocation scheme, and its corresponding total allocation cost is then calculated. This closed-loop process of scheme generation and cost evaluation is continuously iterated, simultaneously accumulating performance evaluation data while generating new candidate solutions. After a preset number of iterations, such as 3000, a total allocation cost set containing a large number of candidate schemes and their precisely quantified costs is finally formed. The preset number of iterations is dynamically determined based on the algorithm's convergence characteristics and computational efficiency requirements. The obtained total allocation cost set completely records the iterative search trajectory, providing a sufficient and reliable decision-making data foundation for subsequent optimal scheme selection based on global comparison.

[0096] Finally, the optimal solution selection is performed. Once the iteration process meets the termination condition, all candidate solutions in the total allocation cost set are evaluated. By comparing the total allocation cost values ​​of each solution, the solution with the minimum total cost is selected as the final target traffic flow allocation solution, representing the allocation strategy that achieves the best balance between safety and economy within the explored solution space.

[0097] In summary, the embodiments of this application have at least the following technical effects:

[0098] Compared to existing technologies, this application first establishes an analytical framework for the interaction between static airspace risk and dynamic traffic flow, providing a scientific basis for traffic flow allocation by quantifying path design capacity and the inherent risk field of airspace. Secondly, it generates a virtual path-probability matrix using an isolated decision-making approach, accurately characterizing the selection behavior of individual UAVs in a risky environment, laying the foundation for group behavior analysis. Furthermore, through a risk pressure evolution assessment mechanism, it innovatively achieves dynamic simulation of the impact of traffic flow aggregation on the risk field, and generates abandonment probability parameters with early warning capabilities.

[0099] Ultimately, through a multi-factor collaborative iterative optimization process, the final allocation scheme not only conforms to the principle of user equilibrium but also effectively controls the overall risk level, achieving a unified optimization of low-altitude traffic flow safety and operational efficiency. This risk evolution-based allocation method can adapt to the complex and ever-changing low-altitude environment, thus providing reliable technical support for large-scale UAV collaborative operations.

[0100] Example 2, as Figure 2 As shown, based on the same inventive concept as the risk-evolution-based low-altitude UAV traffic flow balancing allocation method provided in Embodiment 1, this embodiment of the invention also provides a risk-evolution-based low-altitude UAV traffic flow balancing allocation system, including:

[0101] The airspace topology and risk field acquisition module is used to acquire the airspace traffic flow topology of the target area and determine the path design capacity and inherent airspace risk field of each traffic flow path in the airspace traffic flow topology.

[0102] An isolated path decision module is used to make isolated path decisions for each UAV based on the inherent risk field of the airspace, forming a virtual path-probability matrix.

[0103] The risk pressure assessment and abandonment probability configuration module is used to perform a one-time risk pressure evolution assessment based on the virtual path-probability matrix, and configure the abandonment probability for each traffic flow path according to the risk pressure evolution assessment results.

[0104] The balanced allocation optimization solution module is used to iteratively optimize the balanced allocation of traffic flow using the inherent risk field of the airspace, the virtual path-probability matrix and the rejection probability, and output the converged result as the target traffic flow allocation scheme.

[0105] The spatial topology and risk field acquisition module 11 is specifically used for:

[0106] Obtain the airspace traffic flow topology of the target area, and determine the path design capacity and inherent airspace risk field of each traffic flow path in the airspace traffic flow topology, including:

[0107] Based on the low-altitude traffic planning information of the target area, the airspace traffic flow topology is obtained, and the path design information corresponding to each traffic flow path in the airspace traffic flow topology is matched to extract the path design capacity, path level features and path speed limit features.

[0108] Based on the path level characteristics and the path speed limit characteristics, a preset path risk management plan is queried to obtain the risk management radius corresponding to each traffic flow path.

[0109] Using the risk management radius as the risk field boundary, the neighborhood risk factor set of the airspace traffic flow topology is collected by traversing the area, and the inherent risk field of the airspace is calculated based on the neighborhood risk factor set and the path risk management plan.

[0110] The isolated path decision module 12 is specifically used for:

[0111] Based on the inherent risk field of the airspace, an isolated path decision is made for each UAV, forming a virtual path-probability matrix, including:

[0112] Define the path selection abundance n of the UAV, where n is less than 1;

[0113] Iterate through multiple drones, make isolated path decisions based on the inherent risk field of the airspace, and output each isolated path decision as a candidate path group. Obtain multiple candidate path groups corresponding to multiple drones, wherein each candidate path group includes multiple candidate paths, and each candidate path is associated with a candidate probability.

[0114] Using the path selection abundance n as a probability constraint, traverse multiple candidate path groups, serialize multiple candidate paths in descending order based on the candidate probabilities, perform probability cumulative analysis accordingly, select the top few candidate paths whose probability cumulative analysis results satisfy the probability constraints, and output them as virtual paths.

[0115] The virtual path-probability matrix is ​​constructed by integrating multiple virtual paths with their corresponding alternative probabilities.

[0116] The risk pressure assessment and rejection probability configuration module 13 is specifically used for:

[0117] A single-transaction risk stress evolution assessment is performed based on the virtual path-probability matrix, including:

[0118] Based on the virtual path-probability matrix, the virtual path selection probabilities of all drones are aggregated, and the expected mechanical flow on each traffic flow path is calculated.

[0119] Based on the low-altitude traffic knowledge graph, a dynamic risk factor set is defined, and dynamic risk factor sets of multiple UAVs are collected according to the dynamic risk factor set.

[0120] Based on standard real-world risk data, the standard risk factor set and mechanical standard flow of the traffic flow path are extracted by combining the dynamic risk factor set, and a risk evolution model is constructed accordingly.

[0121] Input the expected mechanical flow and the set of dynamic risk factors into the risk evolution model to obtain the evolved risk field, and output the expected mechanical flow and the evolved risk field as the risk pressure evolution assessment result.

[0122] Furthermore, based on the risk pressure evolution assessment results, an abandonment probability is configured for each of the aforementioned traffic flow paths, including:

[0123] Calculate the risk enhancement coefficient for each traffic flow path, whereby the risk enhancement coefficient is the ratio of the sum of the inherent risk field and the evolved risk field of the airspace to the inherent risk field of the airspace.

[0124] Calculate the flow pressure coefficient for each traffic flow path, where the flow pressure coefficient is the ratio of the expected mechanical flow to the path design capacity;

[0125] The risk enhancement coefficient and the flow pressure coefficient are input into a pre-constructed rejection probability function to calculate the rejection probability of each traffic flow path, wherein the rejection probability function is a monotonically increasing function with respect to the risk enhancement coefficient and the flow pressure coefficient.

[0126] The balanced allocation optimization solution module 14 is specifically used for:

[0127] Using the inherent risk field of the airspace, the virtual path-probability matrix, and the rejection probability, an optimization algorithm is employed to iteratively optimize traffic flow allocation, outputting a converged result as the target traffic flow allocation scheme, including:

[0128] By combining the virtual path-probability matrix with the rejection probability, traffic flow paths are assigned to each UAV in sequence to obtain an initial allocation scheme;

[0129] Based on the inherent risk field of the airspace, the total allocation cost of the initial allocation scheme is calculated and obtained, wherein the total allocation cost includes the total risk cost and the total cost.

[0130] The initial allocation scheme is cross-mutated based on the discard probability, and the total allocation cost set is obtained by iterative calculation based on the cross-mutation result.

[0131] Based on the total allocation cost set, the allocation scheme corresponding to the one with the smallest total allocation cost is selected as the target traffic flow allocation scheme.

[0132] Specifically, by combining the virtual path-probability matrix and the rejection probability, traffic flow paths are assigned to each UAV in sequence to obtain an initial allocation scheme, including:

[0133] Obtain the virtual path and alternative probabilities of the target UAV from the virtual path-probability matrix;

[0134] Based on the aforementioned candidate probabilities, one of the multiple virtual paths of the target UAV is randomly sampled as the first candidate path;

[0135] Generate a random number within a predetermined range and compare the random number with the rejection probability corresponding to the first candidate path;

[0136] When the random number is greater than the rejection probability, the first candidate path is stored as the traffic flow path of the target UAV in the initial allocation scheme.

[0137] When the random number is less than or equal to the rejection probability, the first candidate path is abandoned and random sampling is performed again;

[0138] Repeat the operation until traffic flow path assignment is completed for each drone, and obtain the initial assignment scheme.

[0139] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0140] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0141] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A low-altitude unmanned aerial vehicle (UAV) traffic flow equilibrium allocation method based on risk evolution, characterized in that: include: Obtain the airspace traffic flow topology of the target area, and determine the path design capacity and inherent airspace risk field of each traffic flow path in the airspace traffic flow topology. Based on the inherent risk field of the airspace, an isolated path decision is made for each UAV to form a virtual path-probability matrix; A one-time risk pressure evolution assessment is performed based on the virtual path-probability matrix, and an abandonment probability is configured for each traffic flow path according to the risk pressure evolution assessment results. Using the inherent risk field of the airspace, the virtual path-probability matrix, and the rejection probability, an optimization algorithm is used to iteratively optimize the traffic flow allocation and output the converged result as the target traffic flow allocation scheme. Specifically, based on the risk pressure evolution assessment results, a rejection probability is configured for each traffic flow path, including: Calculate the risk enhancement coefficient for each traffic flow path, whereby the risk enhancement coefficient is the ratio of the sum of the inherent risk field and the evolved risk field of the airspace to the inherent risk field of the airspace. Calculate the flow pressure coefficient for each traffic flow path, where the flow pressure coefficient is the ratio of the expected mechanical flow to the designed capacity of the path; The risk enhancement coefficient and the flow pressure coefficient are input into a pre-constructed rejection probability function to calculate the rejection probability of each traffic flow path, wherein the rejection probability function is a monotonically increasing function with respect to the risk enhancement coefficient and the flow pressure coefficient. Specifically, the inherent risk field of the airspace, the virtual path-probability matrix, and the rejection probability are used to iteratively optimize the traffic flow allocation using an optimization algorithm. The converged result is the target traffic flow allocation scheme, including: By combining the virtual path-probability matrix with the rejection probability, traffic flow paths are assigned to each UAV in sequence to obtain an initial allocation scheme; Based on the inherent risk field of the airspace, the total allocation cost of the initial allocation scheme is calculated and obtained, wherein the total allocation cost includes the total risk cost and the total cost. The initial allocation scheme is cross-mutated based on the discard probability, and the total allocation cost set is obtained by iterative calculation based on the cross-mutation result. Based on the total allocation cost set, the allocation scheme corresponding to the one with the smallest total allocation cost is selected as the target traffic flow allocation scheme.

2. The low-altitude unmanned aerial vehicle traffic flow equilibrium allocation method using risk evolution as described in claim 1, characterized in that, Obtain the airspace traffic flow topology of the target area, and determine the path design capacity and inherent airspace risk field of each traffic flow path in the airspace traffic flow topology, including: Based on the low-altitude traffic planning information of the target area, the airspace traffic flow topology is obtained, and the path design information corresponding to each traffic flow path in the airspace traffic flow topology is matched to extract the path design capacity, path level features and path speed limit features. Based on the path level characteristics and the path speed limit characteristics, a preset path risk management plan is queried to obtain the risk management radius corresponding to each traffic flow path. Using the risk management radius as the risk field boundary, the neighborhood risk factor set of the airspace traffic flow topology is collected by traversing the area, and the inherent risk field of the airspace is calculated based on the neighborhood risk factor set and the path risk management plan.

3. The low-altitude unmanned aerial vehicle traffic flow equilibrium allocation method using risk evolution as described in claim 2, characterized in that, Based on the inherent risk field of the airspace, an isolated path decision is made for each UAV, forming a virtual path-probability matrix, including: Define the path selection abundance n of the UAV, where n is less than 1; Iterate through multiple drones, make isolated path decisions based on the inherent risk field of the airspace, and output each isolated path decision as a candidate path group. Obtain multiple candidate path groups corresponding to multiple drones, wherein each candidate path group includes multiple candidate paths, and each candidate path is associated with a candidate probability. Using the path selection abundance n as a probability constraint, traverse multiple candidate path groups, serialize multiple candidate paths in descending order based on the candidate probabilities, perform probability cumulative analysis accordingly, select the top few candidate paths whose probability cumulative analysis results satisfy the probability constraints, and output them as virtual paths. The virtual path-probability matrix is ​​constructed by integrating multiple virtual paths with their corresponding alternative probabilities.

4. The low-altitude unmanned aerial vehicle traffic flow equilibration allocation method using risk evolution as described in claim 3, characterized in that, A single-transaction risk stress evolution assessment is performed based on the virtual path-probability matrix, including: Based on the virtual path-probability matrix, the virtual path selection probabilities of all drones are aggregated, and the expected mechanical flow on each traffic flow path is calculated. Based on the low-altitude traffic knowledge graph, a dynamic risk factor set is defined, and dynamic risk factor sets of multiple UAVs are collected according to the dynamic risk factor set. Based on standard real-world risk data, the standard risk factor set and mechanical standard flow of the traffic flow path are extracted by combining the dynamic risk factor set, and a risk evolution model is constructed accordingly. Input the expected mechanical flow and the set of dynamic risk factors into the risk evolution model to obtain the evolved risk field, and output the expected mechanical flow and the evolved risk field as the risk pressure evolution assessment result.

5. The low-altitude unmanned aerial vehicle traffic flow equilibrium allocation method using risk evolution as described in claim 1, characterized in that, Combining the virtual path-probability matrix with the rejection probability, traffic flow paths are sequentially assigned to each UAV to obtain an initial allocation scheme, including: Obtain the virtual path and alternative probabilities of the target UAV from the virtual path-probability matrix; Based on the aforementioned candidate probabilities, one of the multiple virtual paths of the target UAV is randomly sampled as the first candidate path. Generate a random number within a predetermined range and compare the random number with the rejection probability corresponding to the first candidate path; When the random number is greater than the rejection probability, the first candidate path is stored as the traffic flow path of the target UAV in the initial allocation scheme. When the random number is less than or equal to the rejection probability, the first candidate path is abandoned and random sampling is performed again; Repeat the operation until traffic flow path assignment is completed for each drone, and obtain the initial assignment scheme.

6. A low-altitude unmanned aerial vehicle (UAV) traffic flow equalization and allocation system employing risk evolution, characterized in that: The method for implementing the risk evolution-based low-altitude unmanned aerial vehicle traffic flow balancing assignment method according to any one of claims 1-5 includes: The airspace topology and risk field acquisition module is used to acquire the airspace traffic flow topology of the target area and determine the path design capacity and inherent airspace risk field of each traffic flow path in the airspace traffic flow topology. An isolated path decision module is used to make isolated path decisions for each UAV based on the inherent risk field of the airspace, forming a virtual path-probability matrix. The risk pressure assessment and abandonment probability configuration module is used to perform a one-time risk pressure evolution assessment based on the virtual path-probability matrix, and configure the abandonment probability for each traffic flow path according to the risk pressure evolution assessment results. The balanced allocation optimization solution module is used to iteratively optimize the balanced allocation of traffic flow using the inherent risk field of the airspace, the virtual path-probability matrix and the rejection probability, and output the converged result as the target traffic flow allocation scheme.

Images