Highway and airspace cooperative low-altitude aircraft route planning method

Abstract

The invention discloses an expressway and airspace cooperative low-altitude aircraft route planning method, and relates to the field of path planning, and the method comprises the steps: optimizing the classification precision of each airspace hierarchy based on a decision tree algorithm and a Bayesian network; repulsive force field parameters are updated through a dynamic situation field updating mechanism; in each airspace level, calculating a gravitational field, a static repulsive force field and a dynamic repulsive force field of the expressway, and fusing to obtain a comprehensive potential field value of each airspace level; acquiring an initial path of an aircraft route by utilizing a greedy strategy and a priority rule and combining the comprehensive potential field value of the airspace level; and according to an artificial potential field local optimization algorithm and a dynamic conflict detection mechanism, adjusting the initial path of the aircraft route to obtain an aircraft route path. According to the invention, an artificial potential field method is combined, a layered collaborative planning algorithm is provided based on airspace ground collaborative potential field modeling, and efficient and safe navigation of the low-altitude aircraft is realized.

Description

Technical Field

[0001] The present invention relates to the field of path planning, and more specifically, to a method for planning flight routes of low-altitude aircraft in coordination with highways and airspace. Background Art

[0002] With the continuous development of low-altitude aircraft, the low-altitude economy has become a strategic emerging economic industry. Deploying low-altitude aircraft using the road network resources of highways can not only share infrastructure, reduce costs, but also improve transportation efficiency and promote regional economic development. However, current methods for planning flight routes of low-altitude aircraft do not consider the topological structure of highways, such as vehicle traffic flow, road network density, and distribution of service facilities.

[0003] Traditional artificial potential field methods have problems such as local minimum traps and poor adaptability to dynamic obstacles in path planning, and do not fully utilize the collaborative relationship between ground infrastructure (such as highways) and low-altitude airspace. Existing technologies are difficult to efficiently generate flight routes that take into account both safety and efficiency in complex airspace environments.

[0004] Specifically, the existing methods for planning flight routes of low-altitude aircraft have the following problems:

[0005] (1) Isolated airspace planning: Traditional low-altitude flight routes rely on airspace obstacle information and do not utilize the navigation value of the ground highway network, resulting in redundant paths and wasted resources.

[0006] (2) Poor dynamic adaptability: Existing potential field methods have insufficient modeling of dynamic obstacles (such as other aircraft, sudden weather areas), and are prone to falling into local minima or obstacle avoidance failure.

[0007] (3) Do not fully utilize the collaborative relationship between the highway road network and low-altitude airspace, resulting in low efficiency and high conflict risk in the planning results.

[0008] In response to the problems in the related art, no effective solution has been proposed yet. Summary of the Invention

[0009] In response to the problems in the related art, the present invention proposes a method for planning flight routes of low-altitude aircraft in coordination with highways and airspace to overcome the above-mentioned technical problems existing in the existing related technologies.

[0010] To this end, the specific technical solution adopted by the present invention is as follows:

[0011] A method for planning flight routes of low-altitude aircraft in coordination with highways and airspace, comprising:

[0012] Based on the decision tree algorithm, conduct a preliminary classification of airspace levels and extract rules, and establish a Bayesian network model containing the probabilistic dependence relationships between airspace levels through a Bayesian network to optimize the classification accuracy between each airspace level;

[0013] Through the dynamic potential field update mechanism, update the repulsive field parameters; within each airspace level, calculate the highway gravitational field, static repulsive field, and dynamic repulsive field, and fuse them to obtain the comprehensive potential field value of each airspace level;

[0014] Utilize the greedy strategy and priority rules, and combine with the comprehensive potential field value of the airspace level to obtain the initial path of the aircraft route;

[0015] According to the artificial potential field local optimization algorithm and dynamic conflict detection mechanism, adjust the initial path of the aircraft route to obtain the aircraft route path.

[0016] Furthermore, based on the decision tree algorithm, conduct a preliminary classification of airspace levels and extract rules, and establish a Bayesian network model containing the probabilistic dependence relationships between airspace levels through a Bayesian network to optimize the classification accuracy between each airspace level includes:

[0017] Obtain the data feature set of the original airspace, including airspace physical features, meteorological features, aircraft operation features, and airspace management features;

[0018] Use the decision tree algorithm to recursively partition the data based on the features in the data feature set, generate a decision tree model, and use the cross-validation method to evaluate the performance of the decision tree model to optimize the decision tree model;

[0019] Predict the new data feature set of the original airspace through the decision tree model to achieve a preliminary airspace level classification; traverse each path of the decision tree model, from the root node to the leaf node, and extract classification rules in the form of conditional relationships;

[0020] Based on the classification rules in the form of conditional relationships, and combined with the conditional probability table calculated by the maximum likelihood estimation method, construct a Bayesian network model containing the probabilistic dependence relationships between airspace levels;

[0021] Input the preliminary classification results into the Bayesian network model, and adjust the preliminary airspace level classification results according to the probabilistic dependence relationships in the Bayesian network.

[0022] Furthermore, traverse each path of the decision tree model, from the root node to the leaf node, and extract classification rules in the form of conditional relationships includes:

[0023] Start from the root node of the decision tree model, traverse each node of the decision tree model through a recursive function until reaching the leaf node; each non-leaf node contains conditions, and each leaf node contains a class label;

[0024] During the recursive process, record all the conditions on each path, and when reaching the leaf node, record the category label corresponding to the leaf node to form a classification rule in the form of a conditional relationship;

[0025] After forming the classification rule in the form of a conditional relationship, it further includes:

[0026] Merge similar classification rules to reduce the complexity of the rule set; apply pruning techniques to remove the overfitting part in the classification rules.

[0027] Furthermore, based on the classification rule in the form of a conditional relationship and combined with the conditional probability table calculated by the maximum likelihood estimation method, constructing a Bayesian network model that includes the probability dependence relationship between airspace levels includes:

[0028] Obtain the classification rule in the form of a conditional relationship, define the variable nodes of the Bayesian network according to the conditions and conclusions in the classification rule in the form of a conditional relationship, and define the edges between the variable nodes according to the causal relationship in the classification rule in the form of a conditional relationship;

[0029] Use the maximum likelihood estimation to calculate the conditional probability table for each variable node and fill the conditional probability table for each variable node;

[0030] Through the variable nodes, the edges between the variable nodes and the conditional probability table, fuse to obtain a Bayesian network model that includes the probability dependence relationship between airspace levels.

[0031] Furthermore, through the dynamic potential field update mechanism, updating the repulsive force field parameters includes:

[0032] According to the urgency of the aircraft mission, adjust the gravitational strength coefficient when calculating the highway gravitational field; according to the endurance time of the aircraft mission, adjust the attenuation radius when calculating the highway gravitational field.

[0033] Furthermore, calculate the highway gravitational field, the static repulsive force field and the dynamic repulsive force field, and fuse to obtain the comprehensive potential field of each airspace level, including:

[0034] Calculate the highway gravitational field strength through the highway gravitational field calculation formula;

[0035] Obtain the geometric center coordinates, height and boundary information of the static obstacles, and divide the airspace into three-dimensional grids; calculate the minimum distance from the center of each three-dimensional grid to all static obstacles;

[0036] When the minimum distance from the center of the three-dimensional grid to all static obstacles is less than or equal to the static obstacle safety distance, then use the static repulsive force field calculation formula to calculate the static repulsive force value strength of the static obstacles; when there is an overlapping area of multiple static obstacles, take the maximum value among the multiple static repulsive force values as the final static repulsive force value strength;

[0037] Obtain the position, speed, heading, and predicted duration of the dynamic obstacle, and calculate the minimum distance between the aircraft and the predicted trajectory of the dynamic obstacle;

[0038] Use the dynamic repulsive force field calculation formula to calculate the intensity of the dynamic repulsive force of the dynamic obstacle, and superimpose it on the airspace grid. At the same time, decay the dynamic repulsive force field of the outdated dynamic obstacle to zero;

[0039] By weighted summing the highway gravitational field intensity, static repulsive force field intensity, and dynamic repulsive force field intensity, obtain the comprehensive potential field value of each airspace level.

[0040] Furthermore, the highway gravitational field calculation formula is:

[0041]

[0042] The static repulsive force field calculation formula is:

[0043]

[0044] The dynamic repulsive force field calculation formula is:

[0045]

[0046] In the formula, d(x, y) is the horizontal distance from the aircraft to the center line of the lane;

[0047] ΔU i is the gravitational increment of the i-th key node of the highway, r i is the influence radius of the i-th key node of the highway;

[0048] N is the total number of key nodes of the highway;

[0049] P is the current three-dimensional position coordinate of the aircraft, P i is the three-dimensional coordinate of the i-th key node of the highway;

[0050] k road is the gravitational strength coefficient, σ is the decay radius;

[0051] d obs is the nearest three-dimensional Euclidean distance from the current position of the aircraft to the surface of the obstacle;

[0052] d safe is the safety distance of the static obstacle;

[0053] k static is the static repulsive force strength coefficient;

[0054] k dvnamic is the dynamic repulsive force strength coefficient;

[0055] v rel is the modulus length difference between the speed vectors of the aircraft and the dynamic obstacle;

[0056] d proj is the minimum distance between the predicted trajectory of the aircraft and the dynamic obstacle;

[0057] t obs is the obstacle state update timestamp; τ is the time decay constant;

[0058] x, y, and z are the three-dimensional space coordinates representing the current position of the aircraft, and t is the current timestamp.

[0059] Furthermore, using the greedy strategy and priority rules, and combining with the comprehensive potential field value of the airspace hierarchy, obtaining the initial path of the aircraft route includes:

[0060] Generating a path skeleton through the path nodes in the highway gravitational field, and using the A* algorithm to connect the path nodes to obtain a preliminary path;

[0061] Starting from the starting point of the aircraft, by iteratively selecting the path node with the lowest comprehensive potential field value in the adjacent three-dimensional grids, and combining with the pre-set priority rules, after optimizing the preliminary path, obtaining the initial path of the aircraft route.

[0062] Furthermore, according to the artificial potential field local optimization algorithm and the dynamic conflict detection mechanism, adjusting the initial path of the aircraft route to obtain the aircraft route path includes:

[0063] Taking the principle of gradually decreasing the comprehensive potential field value, using the gradient descent method to iteratively adjust the path node positions in the initial path of the aircraft route;

[0064] Dividing the airspace into spatio-temporal grids, and predicting the future trajectory occupancy of the aircraft and the obstacles;

[0065] When there are multiple aircraft trajectories in the same spatio-temporal grid, and the predicted distance between the aircraft trajectory and the obstacle is less than the static obstacle safety distance, it is determined that the current path of the aircraft route is in a conflict state, and according to the pre-set dynamic adjustment strategy, adjusting the current path of the aircraft route to obtain the aircraft route path.

[0066] Furthermore, taking the principle of gradually decreasing the comprehensive potential field value, using the gradient descent method to iteratively adjust the path node positions in the initial path of the aircraft route includes:

[0067] Using the node position adjustment formula to adjust the path node positions in the initial path of the aircraft route, and the node position adjustment formula:

[0068]

[0069] In the formula, P i+1 is the three-dimensional coordinate of the path node at the (i + 1)-th iteration, and P i is the three-dimensional coordinate of the path node at the i-th iteration;

[0070] η is the learning rate; is the gradient vector of the comprehensive potential field at the node P i .

[0071] The beneficial effects of the present invention are as follows:

[0072] (1) Physical-information space coupling: By integrating the highway and airspace potential fields, the ground infrastructure is transformed into airspace navigation resources, reducing flight energy consumption.

[0073] (2) Multi-granularity obstacle modeling: The method of precise modeling of static obstacles + real-time prediction of dynamic obstacles is adopted, which can balance safety and computational efficiency.

[0074] (3) The present invention improves the artificial potential field model, introduces the cooperative action of the highway gravitational field and the dynamic obstacle repulsive field, and realizes the rapid generation and dynamic optimization of the low-altitude aircraft flight path.

[0075] (4) The present invention gives full play to the synergy between the low-altitude aircraft flight path planning and the ground traffic infrastructure, effectively improving the rationality and efficiency of the low-altitude aircraft flight path.

[0076] (5) The present invention is based on the combined method of decision tree and Bayesian network, which not only optimizes the airspace level classification in the low-altitude aircraft flight path planning, but also simplifies the rule set, enhances the interpretability and practicability of the model, provides strong support for the low-altitude aircraft flight path planning, makes the altitude adjustment more accurate and flexible during the flight path of the aircraft, and is applicable to the specific external environment and aircraft mission requirements of the aircraft at present. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0078] Figure 1 is a flowchart of a method for planning a low-altitude aircraft flight path with highway-airspace cooperation according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0079] To further illustrate the embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be combined with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, those of ordinary skill in the art should be able to understand other possible implementation manners and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are usually used to represent similar components.

[0080] According to an embodiment of the present invention, there is provided a method for planning a flight route of a low-altitude aircraft in coordination with a highway and airspace.

[0081] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments. As Figure 1 shown, the method for planning a flight route of a low-altitude aircraft in coordination with a highway and airspace according to an embodiment of the present invention includes:

[0082] S1. Based on the decision tree algorithm, conduct preliminary classification and rule extraction for airspace levels, and establish a Bayesian network model including the probability dependence relationship between airspace levels through a Bayesian network to optimize the classification accuracy between airspace levels.

[0083] S2. Through the dynamic potential field update mechanism, update the repulsive field parameters; within each airspace level, calculate the highway gravitational field, static repulsive field, and dynamic repulsive field, and fuse them to obtain the comprehensive potential field value of each airspace level.

[0084] S3. Utilize the greedy strategy and priority rules, and combine with the comprehensive potential field value of the airspace level to obtain the initial path of the aircraft flight route.

[0085] S4. According to the artificial potential field local optimization algorithm and dynamic conflict detection mechanism, adjust the initial path of the aircraft flight route to obtain the aircraft flight route.

[0086] In one embodiment, based on the decision tree algorithm, conducting preliminary classification and rule extraction for airspace levels, and establishing a Bayesian network model including the probability dependence relationship between airspace levels through a Bayesian network to optimize the classification accuracy between airspace levels includes:

[0087] Obtain the data feature set of the original airspace, including airspace physical features, meteorological features, aircraft operation features, and airspace management features.

[0088] Among them, the airspace physical features: such as geographical coordinates, altitude, etc.

[0089] The meteorological features: such as temperature, humidity, wind speed, etc.

[0090] The aircraft operation features: such as speed, heading, load capacity, etc.

[0091] Airspace management features: such as control area boundaries, flight restricted areas, etc.

[0092] Use the decision tree algorithm to recursively partition the data based on the features in the data feature set, generate a decision tree model, and use the cross-validation method to evaluate the performance of the decision tree model to optimize the decision tree model.

[0093] Use the decision tree model to predict the new data feature set of the original airspace to achieve preliminary airspace level classification; traverse each path of the decision tree model, and extract classification rules in the form of conditional relationships from the root node to the leaf node.

[0094] Based on the classification rules in the form of conditional relationships and combined with the conditional probability table calculated by the maximum likelihood estimation method, construct a Bayesian network model that includes the probability dependence relationship between airspace levels.

[0095] Input the preliminary classification result into the Bayesian network model, and adjust the preliminary airspace level classification result according to the probability dependence relationship in the Bayesian network.

[0096] It should be noted that the decision tree algorithm is used to recursively partition the data based on the information in the above-mentioned feature set. The goal is to generate a decision tree model that can predict the corresponding airspace level category according to the input features. To ensure the effectiveness of the model and avoid overfitting, the cross-validation method is usually used to evaluate and optimize the model performance.

[0097] Use the generated decision tree model to predict the new airspace data to achieve preliminary airspace level classification. At the same time, traverse each path of the decision tree (from the root node to the leaf node), and extract classification rules in the form of conditional relationships (IF-THEN) from it. These rules essentially describe how to determine the airspace level attribution under specific conditions.

[0098] Based on the obtained classification rules and the conditional probability table calculated by the maximum likelihood estimation method, construct a Bayesian network model. This model not only includes the causal relationship between variable nodes (defined by the classification rules), but also considers the quantitative representation of the probability dependence between variables (through the conditional probability table).

[0099] Take the preliminary classification result obtained through the decision tree model as input and apply it to the constructed Bayesian network model. According to the probability dependence relationship in the Bayesian network, further adjust and optimize the initial airspace level classification result. This can more accurately reflect the uncertainty factors in the actual situation and improve the classification accuracy.

[0100] In one embodiment, traversing each path of the decision tree model, the classification rules in the form of conditional relationships extracted from the root node to the leaf node include:

[0101] Starting from the root node of the decision tree model, traverse each node of the decision tree model through a recursive function until reaching the leaf node; each non-leaf node contains a condition, and each leaf node contains a class label.

[0102] During the recursive process, record all the conditions on each path, and when reaching the leaf node, record the class label corresponding to the leaf node to form a classification rule in the form of a conditional relationship. For example, "If the temperature < 20 degrees and the humidity > 80%, then it is airspace level A".

[0103] After forming the classification rule in the form of a conditional relationship, it also includes:

[0104] Merge similar classification rules to reduce the complexity of the rule set; apply pruning techniques to remove the overfitting part in the classification rules.

[0105] A recursive function is a function that directly or indirectly calls itself in its definition or implementation. During the process of traversing the decision tree model, the recursive function records all the conditions on the path by gradually delving into each branch from the root node until reaching the leaf node, and finally forms a classification rule.

[0106] Define a recursive function extract_rules that takes two parameters: the currently processed node and the list of conditions (conditions) accumulated so far. At the initial call, conditions is an empty list. This list grows dynamically during the recursive process. Each time a new branch is entered, the condition of that branch is added to the list. When reaching the leaf node, the accumulated list of conditions contains all the conditions on the path from the root node to that leaf node.

[0107] Through pruning techniques (such as pre-pruning or post-pruning), remove to a certain extent those parts that contribute little to the overall classification but increase the model complexity. In the rule extraction stage, consider removing those overly specific or rules generated based on a small number of samples, thereby improving the stability and accuracy of the rule set.

[0108] In one embodiment, based on the classification rule in the form of a conditional relationship and combined with the conditional probability table calculated by the maximum likelihood estimation method, constructing a Bayesian network model including the probability dependence relationship between airspace levels includes:

[0109] Obtain the classification rule in the form of a conditional relationship, define the variable nodes of the Bayesian network according to the conditions and conclusions in the classification rule in the form of a conditional relationship, and define the edges between the variable nodes according to the causal relationship in the classification rule in the form of a conditional relationship.

[0110] Use the maximum likelihood estimation to calculate the conditional probability table for each variable node and fill the conditional probability table for each variable node.

[0111] A Bayesian network model containing the probability dependence relationship between spatial domain levels is obtained through variable nodes, the edges between variable nodes, and conditional probability tables.

[0112] For example, "temperature < 20 degrees", "humidity > 80%", etc. can be used as variable nodes. "Temperature < 20 degrees" and "humidity > 80%" result in "spatial domain level A", which indicates a causal relationship, thus establishing a directed edge between the corresponding variable nodes.

[0113] Maximum likelihood estimation is a parameter estimation method used to estimate the best values of model parameters for a given data set. That is, the probability that the node takes different states when the states of its parent nodes are given.

[0114] First, collect data: First, a sufficiently large data set containing historical observations of all relevant variables is required.

[0115] Calculate probabilities: For each variable node, based on the state combinations of its parent nodes, calculate the probabilities that the node is in each possible state. Specifically, it is to count the frequencies of the child nodes in various states under different parent node states, and then convert these frequencies into probabilities.

[0116] For example, there is a variable node "spatial domain level A", whose parent nodes are "temperature < 20 degrees" and "humidity > 80%". Calculate the probability that "spatial domain level A = true" when "temperature < 20 degrees = true" and "humidity > 80% = true"; and the probabilities that "spatial domain level A = true" or "false" when these two conditions are in other combinations. This is done by counting the number of instances in the training data set that meet these conditions and dividing by the total number.

[0117] In one embodiment, through a dynamic potential field update mechanism, updating the repulsive force field parameters includes:

[0118] According to the urgency of the aircraft mission, adjust the gravitational strength coefficient when calculating the highway gravitational field.

[0119] According to the endurance duration of the aircraft mission, adjust the attenuation radius when calculating the highway gravitational field.

[0120] In one embodiment, calculating the highway gravitational field, static repulsive force field, and dynamic repulsive force field, and fusing them to obtain the comprehensive potential field of each spatial domain level includes:

[0121] Calculate the highway gravitational field strength through the highway gravitational field calculation formula.

[0122] Obtain the geometric center coordinates, height, and boundary information of static obstacles, and divide the airspace into three-dimensional grids; calculate the minimum distance from the center of each three-dimensional grid to all static obstacles.

[0123] When the minimum distance from the center of the three-dimensional grid to all static obstacles is less than or equal to the static obstacle safety distance, the static repulsive force value intensity of the static obstacle is calculated using the static repulsive force field calculation formula; when there is an overlapping area of multiple static obstacles, the maximum value among the multiple static repulsive force values is taken as the final static repulsive force value intensity.

[0124] Obtain the position, velocity, heading, and predicted duration of the dynamic obstacle, and calculate the minimum distance between the aircraft and the predicted trajectory of the dynamic obstacle.

[0125] Use the dynamic repulsive force field calculation formula to calculate the dynamic repulsive force value intensity of the dynamic obstacle and superimpose it on the airspace grid. At the same time, attenuate the dynamic repulsive force field of the outdated dynamic obstacle to zero.

[0126] By weighted summing the highway gravitational field intensity, static repulsive force field intensity, and dynamic repulsive force field intensity, the comprehensive potential field value of each airspace level is obtained.

[0127] In one embodiment, the highway gravitational field calculation formula is:

[0128]

[0129] The static repulsive force field calculation formula is:

[0130]

[0131] The dynamic repulsive force field calculation formula is:

[0132]

[0133] In the formula, d(x, y) is the horizontal distance from the aircraft to the center line of the lane; ΔU i is the gravitational increment of the i-th key highway node; r i is the influence radius of the i-th key highway node; N is the total number of key highway nodes; P is the current three-dimensional position coordinate of the aircraft, P i is the three-dimensional coordinate of the i-th key highway node; k road is the gravitational strength coefficient, σ is the attenuation radius; d obs is the nearest three-dimensional Euclidean distance from the current position of the aircraft to the surface of the obstacle; d safe is the static obstacle safety distance; k static is the static repulsive force strength coefficient; k dynamic is the dynamic repulsive force strength coefficient; v rel is the modulus difference of the velocity vectors between the aircraft and the dynamic obstacle; d proj is the minimum distance between the aircraft and the predicted trajectory of the dynamic obstacle; t obsis the timestamp for obstacle state update; τ is the time decay constant; x, y, and z are the three-dimensional spatial coordinates representing the current position of the aircraft, and t is the current timestamp.

[0134] In one embodiment, using the greedy strategy and priority rules, and combining with the comprehensive potential field value of the airspace hierarchy, obtaining the initial path of the aircraft route includes:

[0135] Generate a path skeleton through the path nodes in the highway gravitational field, and use the A* algorithm to connect the path nodes to obtain a preliminary path.

[0136] Starting from the starting point of the aircraft, by iteratively selecting the path node with the lowest comprehensive potential field value in the adjacent three-dimensional grids, and combining with the pre-set priority rules, after optimizing the preliminary path, obtain the initial path of the aircraft route.

[0137] In one embodiment, according to the artificial potential field local optimization algorithm and the dynamic conflict detection mechanism, adjust the initial path of the aircraft route to obtain the aircraft route path, including:

[0138] Taking the principle of gradually decreasing the comprehensive potential field value, using the gradient descent method, iteratively adjust the position of the path nodes in the initial path of the aircraft route.

[0139] Divide the airspace into spatio-temporal grids, and predict the future trajectory occupancy of the aircraft and obstacles.

[0140] When there are multiple aircraft trajectories in the same spatio-temporal grid, and the predicted distance between the aircraft trajectory and the obstacle is less than the static obstacle safety distance, it is determined that the current path of the aircraft route is in a conflict state, and according to the pre-set dynamic adjustment strategy, adjust the current path of the aircraft route to obtain the aircraft route path.

[0141] In one embodiment, taking the principle of gradually decreasing the comprehensive potential field value, using the gradient descent method, iteratively adjusting the position of the path nodes in the initial path of the aircraft route includes:

[0142] Use the node position adjustment formula to adjust the position of the path nodes in the initial path of the aircraft route, and the node position adjustment formula:

[0143]

[0144] In the formula, P i+1 is the three-dimensional coordinate of the path node at the (i + 1)-th iteration, P i is the three-dimensional coordinate of the path node at the i-th iteration; η is the learning rate; is the gradient vector of the comprehensive potential field at the node P i .

[0145] To facilitate the understanding of the above technical solution of the present invention, the working principle of the present invention in the actual process will be described in detail below.

[0146] I: Airspace - Highway Cooperative Data Collection and Preprocessing

[0147] 1. Data Input

[0148] (1) Ground Data

[0149] The GIS coordinates of the center line of the highway lanes and the three - dimensional position information of key nodes (service areas, overpasses, etc.).

[0150] (2) Airspace Data

[0151] The three - dimensional Bounding Box or point cloud data of static obstacles (buildings, mountains, etc.) and the real - time states (position, speed, heading, etc.) of dynamic obstacles (other aircraft, meteorological areas, etc.).

[0152] (3) Airspace Control Information

[0153] The positions and valid times of no - fly zones, temporary control zones, etc.

[0154] II: Airspace - Ground Cooperative Potential Field Modeling

[0155] 1. Generation of Highway Gravitational Field

[0156] (1) Gravitational Field Source: The center line of the highway lanes is mapped into low - altitude gravitational lines, and the gravitational intensity decays with distance. The formula for the gravitational field is:

[0157]

[0158] Among them, d(x, y) is the horizontal distance from the aircraft to the center line of the lane; ΔU i is the gravitational increment of the i - th node, generally 100 ≤ ΔU i ≤ 300; r i is the influence radius of the i - th node, generally 200m ≤ r i ≤ 800m; N is the total number of key nodes (service areas, overpasses, etc.) on the highway; P=(x, y, z) is the current three - dimensional position coordinates of the aircraft; P i is the three - dimensional coordinates of the i - th key node (such as a service area); k road is the gravitational intensity coefficient, which controls the attraction of the center line of the highway lane to the aircraft. The value range is generally 100 ≤ k road ≤ 200. The larger the value, the stronger the gravitational force; σ is the decay radius, which controls the decay rate of the gravitational field with distance. The value range is usually 100m ≤ σ ≤ 500m.

[0159] 2. Static Repulsive Force Field Modeling

[0160] (1) The static repulsive force field is used to model fixed obstacles (such as buildings, mountains, high-voltage towers, etc.). Its intensity is inversely proportional to the distance from the obstacle and takes effect within the safety distance. The calculation formula for the static repulsive force field is:

[0161]

[0162] where d obs is the nearest three-dimensional Euclidean distance (unit: meter) from the current position of the aircraft to the surface of the obstacle;

[0163] d safe is the safety distance of the static obstacle, which is positively correlated with the height of the obstacle. The formula is:

[0164] d safe = α·h obs + β;

[0165] h obs is the height of the obstacle; α is the height influence coefficient, generally taking values in the range of 0.3 - 0.5. The higher the building height, the larger the value; β is the basic safety distance, generally taking values in the range of 20 - 50m; k static is the static repulsive force intensity coefficient, generally taking values of 100 - 500, which is adjusted according to the danger level of the obstacle.

[0166] (2) Technical Implementation Process

[0167] 1) Obstacle Data Input

[0168] Load the three-dimensional model (Bounding Box or point cloud data) of the obstacle from the airspace control database.

[0169] Extract the geometric center coordinates, height, and boundary information of the obstacle.

[0170] 2) Spatial Discretization

[0171] Divide the airspace into three-dimensional grids (such as 50m × 50m × 20m), and calculate the minimum distance d obs .

[0172] 3) Repulsive Force Field Assignment

[0173] If d obs ≤ d safe , calculate the repulsive force value according to the static repulsive force field calculation formula.

[0174] For the overlapping area of multiple obstacles, take the maximum value as the final repulsive force field intensity.

[0175] 3. Dynamic Repulsive Force Field Modeling

[0176] The dynamic repulsive force field is used to model time-varying obstacles (such as other aircraft, sudden meteorological areas), and its intensity is related to the motion state of the obstacle and the predicted risk. The calculation formula of the dynamic repulsive force field is as follows:

[0177]

[0178] Among them, k dynamic is the dynamic repulsive force intensity coefficient, which controls the overall amplitude of the repulsive force field, and the value range is generally 200 - 800; v rel is the relative velocity, which is the modulus difference between the velocity vectors of the aircraft and the obstacle; d proj is the predicted minimum approach distance, which is the minimum distance between the aircraft and the obstacle predicted in the future t pred seconds. t obs is the timestamp for updating the obstacle state; τ is the time decay constant, which controls the decay rate of the repulsive force field over time. The smaller the value, the faster the decay; x, y, and z are the three-dimensional space coordinates representing the current position of the aircraft; t is the current timestamp.

[0179] Technical implementation process

[0180] (1) Input of dynamic obstacle data:

[0181] Receive ADS - B signals, air traffic control radar data, or meteorological warning information in real time.

[0182] Analyze the position, velocity, heading, and expected duration of the obstacle (such as the movement path of a thunderstorm).

[0183] For example, obtain the latest meteorological data from weather radars, satellite images, and ground observation stations. Input these data into a numerical weather prediction model to generate short - to medium - term weather forecasts. Analyze the results output by the model by meteorological experts or intelligent systems to determine the specific path of the thunderstorm and the time points when it is expected to reach various locations. Based on the above analysis, issue accurate warning information to relevant parties, informing them of the expected impact time of the thunderstorm and the recommended preventive measures. Continuously monitor weather changes throughout the process and adjust the prediction according to the latest data to ensure the accuracy of the information. Obtain a relatively accurate expected duration of the obstacle (such as a thunderstorm) through these technologies, thus providing important support for application scenarios such as low - altitude aircraft route planning.

[0184] (2) Trajectory prediction and risk calculation

[0185] Trajectory prediction: Predict the trajectory within the future t pred seconds (generally 5 - 15S).

[0186] Minimum approach distance: Calculate the minimum distance d proj .

[0187] For example, obtain the latest information on the position, speed, and direction of thunderstorms from the meteorological department, and at the same time obtain the speed, altitude, and planned route of the aircraft from the flight control system. Use a numerical weather prediction model to predict the movement path of thunderstorms in the next few hours. Input the predicted paths of the aircraft and thunderstorms into a calculation program to find the minimum distance between the two.

[0188] (3) Repulsive force field update

[0189] Calculate the dynamic repulsive force value according to the formula and superimpose it on the airspace grid.

[0190] Automatically attenuate the repulsive force field of obsolete obstacles [(t - t obs ) > τ] to zero.

[0191] 4. Cooperative potential field fusion and altitude stratification

[0192] (1) Potential field superposition rule

[0193] Comprehensive potential field calculation:

[0194] U total (x, y, z, t) = w att ·U att + w rep ·(U rep-static + U rep-dynamic );

[0195] Among them, w att , w rep are the gravitational and repulsive force weight coefficients respectively. By default, w att = 1.0, w rep = 1.2.

[0196] (2) Airspace altitude stratification strategy

[0197] Stratification definition:

[0198] Divide the low-altitude airspace into multiple levels according to altitude (for example: 0 - 50m is the no-fly layer, 50 - 100m is the low-altitude logistics layer, 100 - 200m is the high-speed traffic layer).

[0199] Collect the data feature set of the original airspace, train a decision tree model, and use the cross-validation method to evaluate the model performance. For example, use 10-fold cross-validation to ensure the generalization ability of the model, and optimize the model by adjusting parameters (such as maximum depth, minimum sample split number, etc.) to avoid overfitting.

[0200] Recursively traverse each path from the root node until the leaf node, and record the conditions of each non-leaf node and the category label corresponding to the leaf node. For example, "If the current flight altitude > 50 meters and < 100 meters, then it is airspace level B (low-altitude logistics layer)".

[0201] Extraction rule: Based on the traversal results, form specific rules, such as "If the visibility > 5 km and the flight altitude is between 50 - 100 m, it is recommended as the low-altitude logistics layer."

[0202] Merge similar rules: Identify rules with similar conditions but different category labels and attempt to merge them to simplify the rule set.

[0203] Pruning: Remove those branches that are too specific or contribute little to the overall classification to reduce the risk of overfitting.

[0204] According to the extracted classification rules, use each condition (such as visibility, flight altitude, etc.) as a variable node and establish directed edges between the nodes based on the causal relationship. For example, "visibility" affects the decision of whether it is suitable for the "low-altitude logistics layer".

[0205] Calculate the conditional probability table: For each variable node, use the maximum likelihood estimation method to calculate its conditional probability table based on the training data. For example, calculate the probability that "belongs to the low-altitude logistics layer = true" when "visibility > 5 km = true" and "flight altitude is between 50 - 100 m = true".

[0206] Construct the network: Combine all variable nodes, the edges between them, and the corresponding conditional probability tables to construct the final Bayesian network model. This model not only reflects the relationships between variables but also quantifies the probabilistic dependencies of these relationships.

[0207] Input the preliminary classification results into the constructed Bayesian network model and use the probabilistic dependencies in the model to further adjust and refine the classification results of each airspace level. For example, for a newly input flight mission, if the preliminary classification suggests "low-altitude logistics layer", but the Bayesian network analysis shows that there is a high degree of uncertainty under the current meteorological conditions, it may be necessary to re-evaluate or select a safer flight altitude range.

[0208] Stratified potential field calculation:

[0209] Calculate the gravitational and repulsive force fields independently for each layer to avoid cross-layer interference; when the aircraft switches altitude layers, it is necessary to ensure the continuity of the vertical potential field gradient (for example, when descending from the 100 m layer to the 80 m layer, it is necessary to ensure that there are no high-repulsion obstacles below).

[0210] 5. Dynamic potential field update mechanism

[0211] Real-time data access:

[0212] (1) Access real-time meteorological data (wind speed, rainfall), airspace control instructions, and the status of other aircraft through 5G / satellite links.

[0213] (2) Update the repulsive force field parameters of dynamic obstacles (position, velocity, predicted trajectory).

[0214] Adaptive parameter adjustment:

[0215] (1) Automatically adjust k according to the task urgency road , for example, for urgent tasks, k can be increased road so that the gravitational force is strong and the flight path is close to the highway; reducing the value of k road results in a weak gravitational force, allowing a larger deviation from the highway to avoid obstacles.

[0216] (2) Adjust σ according to the energy consumption limit. For example, increase the value of σ during long-endurance tasks, thereby expanding the range of action of the gravitational field and reducing yaw energy consumption.

[0217] III: Initial path generation

[0218] 1. Path skeleton construction

[0219] Select key nodes along the highway gravitational field (such as one node every 5 kilometers) to generate the initial path skeleton; connect the nodes through the A* algorithm, avoiding high-value areas of the static repulsive force field. The A* algorithm, also known as the A-star algorithm, is a type of heuristic search algorithm.

[0220] 2. Heuristic path expansion

[0221] Greedy strategy: Starting from the starting point, iteratively select the node with the lowest comprehensive potential field value in the adjacent grid.

[0222] Priority rule: Aircraft for urgent tasks prefer to use the low-altitude layer (50 - 100 meters), and aircraft for long-endurance tasks prefer to fly along the highway gravitational field.

[0223] IV: Path dynamic optimization and conflict resolution

[0224] 1. Artificial potential field local optimization

[0225] Gradient descent method: Iteratively adjust the node positions along the initial path to gradually reduce the path potential field value; the node position adjustment formula:

[0226]

[0227] where P i+1 is the three-dimensional coordinate of the path node at the (i + 1)-th iteration, and P i is the three-dimensional coordinate of the path node at the i-th iteration. η is the learning rate (step size factor). is the gradient vector of the comprehensive potential field at the node P i , representing the direction rate of the fastest change in the potential field.

[0228] 2. Dynamic conflict detection

[0229] Space-time cube model: Divide the airspace into space-time grids (for example: spatial grid 50m×50m×20m, time resolution 1 - 5 seconds); predict the trajectory occupancy of the aircraft and obstacles within the next 30s.

[0230] For example, obtain the positions of the aircraft and obstacles at each time point. Record these positions to form a trajectory set. After knowing the positions of the aircraft and obstacles, the distance between the aircraft and the obstacles can be calculated. If the distance between the aircraft and the obstacles is less than the safety threshold, it is considered that there is a potential conflict.

[0231] Conflict determination conditions: There are multiple aircraft trajectories within the same space-time grid; the predicted distance d between the aircraft trajectory and the obstacle proj <d safe 。

[0232] 3. Dynamic adjustment strategy

[0233] Horizontal obstacle avoidance: Shift to the adjacent airspace corridor (preferably choose the corridor with a lower potential field value).

[0234] Vertical obstacle avoidance: Switch to the idle altitude layer (such as climbing from the 100m layer to the 150m layer).

[0235] Temporal sequence obstacle avoidance: Switch to the idle altitude layer (such as climbing from the 100m layer to the 150m layer).

[0236] In summary, the present invention combines the highway and the airspace potential field, transforms the ground infrastructure into airspace navigation resources, and reduces flight energy consumption. By adopting the method of precise modeling of static obstacles + real-time prediction of dynamic obstacles, it can balance safety and computational efficiency. By improving the artificial potential field model and introducing the cooperative action of the highway gravitational field and the dynamic obstacle repulsive field, it realizes the rapid generation and dynamic optimization of the low-altitude aircraft route. It gives full play to the coordination between the low-altitude aircraft route planning and the ground traffic infrastructure, effectively improves the rationality and efficiency of the low-altitude aircraft route. Based on the combined method of decision tree and Bayesian network, it not only optimizes the airspace level classification in the low-altitude aircraft route planning, but also simplifies the rule set, enhances the interpretability and practicability of the model, provides strong support for the low-altitude aircraft route planning, makes the height adjustment more accurate and flexible when the aircraft is on the route, and is applicable to the specific external environment and aircraft mission requirements of the aircraft at present.

[0237] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for low-altitude aircraft route planning in coordination between highway and airspace, characterized in that: include: Based on the decision tree algorithm, the spatial hierarchy is preliminarily classified and rules are extracted, and a Bayesian network model containing the probabilistic dependency relationship between the spatial hierarchy is established through the Bayesian network to optimize the classification accuracy between the spatial hierarchy. Update the repulsive field parameters through the dynamic potential field update mechanism; calculate the highway gravitational field, static repulsive field and dynamic repulsive field in each airspace level, and integrate them to obtain the comprehensive potential field value of each airspace level; Using the greedy strategy and priority rules, combined with the comprehensive potential field value at the airspace level, the initial path of the aircraft route is obtained; According to the artificial potential field local optimization algorithm and the dynamic conflict detection mechanism, the initial path of the aircraft route is adjusted to obtain the aircraft route path.

2. The method for low-altitude aircraft route planning based on highway and airspace coordination according to claim 1 is characterized in that: The method of performing preliminary classification and rule extraction on the airspace levels based on the decision tree algorithm and establishing a Bayesian network model including the probability dependency relationship between airspace levels through the Bayesian network to optimize the classification accuracy between each airspace level includes: Obtaining the data feature set of the original airspace, including airspace physical features, meteorological features, aircraft operation features and airspace management features; Use the decision tree algorithm to recursively divide the data based on the features in the data feature set to generate a decision tree model, and use the cross-validation method to evaluate the performance of the decision tree model to optimize the decision tree model; The new data feature set of the original airspace is predicted through the decision tree model to achieve preliminary airspace hierarchical classification; each path of the decision tree model is traversed, and classification rules in the form of conditional relationships are extracted from the root node to the leaf node; Based on the classification rules in the form of conditional relations and combined with the conditional probability table calculated by the maximum likelihood estimation method, a Bayesian network model containing the probability dependency relationship between spatial hierarchies is constructed; The preliminary classification results are input into the Bayesian network model, and the preliminary spatial hierarchical classification results are adjusted according to the probabilistic dependency relationship in the Bayesian network.

3. The method for low-altitude aircraft route planning based on highway and airspace coordination according to claim 2 is characterized in that: The traversal of each path of the decision tree model, from the root node to the leaf node, extracts the classification rules in the form of conditional relationships, including: Starting from the root node of the decision tree model, a recursive function is used to traverse each node of the decision tree model until a leaf node is reached; each non-leaf node contains a condition, and each leaf node contains a category label; In the recursive process, all conditions on each path are recorded, and when a leaf node is reached, the category label corresponding to the leaf node is recorded to form a classification rule in the form of a conditional relationship; The classification rules that form the conditional relationship also include: Merge similar classification rules to reduce the complexity of the rule set; apply pruning techniques to remove overfitting parts in the classification rules.

4. The method for low-altitude aircraft route planning based on highway and airspace coordination according to claim 2 is characterized in that: The classification rules based on the conditional relationship form and the conditional probability table calculated by the maximum likelihood estimation method are combined to construct a Bayesian network model containing the probability dependency relationship between spatial domain levels, including: Obtaining classification rules in the form of conditional relationships, defining variable nodes of the Bayesian network according to the condition and conclusion parts in the classification rules in the form of conditional relationships, and defining edges between the variable nodes according to the causal relationship in the classification rules in the form of conditional relationships; Use maximum likelihood estimation to calculate the conditional probability table for each variable node and fill the conditional probability table for each variable node; Through the integration of variable nodes, edges between variable nodes and conditional probability tables, a Bayesian network model containing probability dependencies between spatial levels is obtained.

5. The method for low-altitude aircraft route planning based on highway and airspace coordination according to claim 1 is characterized in that: The updating of the repulsive field parameters by the dynamic potential field updating mechanism includes: According to the urgency of the aircraft mission, the gravity intensity coefficient when calculating the highway gravity field is adjusted; Adjust the attenuation radius when calculating the highway gravity field based on the aircraft's mission endurance.

6. The method for low-altitude aircraft route planning in coordination between highway and airspace according to claim 1, characterized in that: The calculation of the highway gravitational field, static repulsive field and dynamic repulsive field, and the fusion of the comprehensive potential field of each airspace level include: Calculate the gravitational field strength of the highway through the calculation formula of the gravitational field of the highway; Obtain the geometric center coordinates, height and boundary information of static obstacles, and divide the airspace into three-dimensional grids; calculate the minimum distance from the center of each three-dimensional grid to all static obstacles; When the minimum distance from the center of the three-dimensional grid to all static obstacles is less than or equal to the static obstacle safety distance, the static repulsion field calculation formula is used to calculate the static repulsion value strength of the static obstacle; when multiple static obstacles overlap, the maximum value of the multiple static repulsion values ​​is taken as the final static repulsion value strength; Obtain the position, speed, heading and estimated duration of the dynamic obstacle, and calculate the minimum distance between the aircraft and the predicted trajectory of the dynamic obstacle; The dynamic repulsion field calculation formula is used to calculate the dynamic repulsion value strength of the dynamic obstacle and superimpose it on the airspace grid. At the same time, the dynamic repulsion field of the outdated dynamic obstacle is attenuated to zero; The comprehensive potential field value of each airspace level is obtained by taking the weighted sum of the gravitational field strength, static repulsive field strength and dynamic repulsive field strength of the highway.

7. The method for low-altitude aircraft route planning based on highway and airspace coordination according to claim 6 is characterized in that: The calculation formula of the highway gravitational field is: The calculation formula of static repulsive field is: The dynamic repulsive field calculation formula is: Where d(x,y) is the horizontal distance from the aircraft to the center line of the lane; ΔU i is the gravitational increment r of the key node of the ith highway i is the influence radius of the key node of the ith highway; N is the total number of key nodes on the highway; P is the current three-dimensional position coordinate of the aircraft, P i is the three-dimensional coordinate of the key node of the i-th highway; k road is the gravitational intensity coefficient, σ is the attenuation radius; d obs is the closest three-dimensional Euclidean distance from the current position of the aircraft to the obstacle surface; d safe is the safety distance for static obstacles; k static is the static repulsion strength coefficient; k dynamic is the dynamic repulsion strength coefficient; v rel is the difference in modulus between the velocity vectors of the aircraft and the dynamic obstacle; d proj The minimum distance between the predicted trajectory of the aircraft and the dynamic obstacle; t obs is the obstacle status update timestamp; τ is the time decay constant; x, y and z are the three-dimensional spatial coordinates representing the current position of the aircraft, and t is the current timestamp.

8. The method for low-altitude aircraft route planning based on highway and airspace coordination according to claim 1 is characterized in that: The method of using the greedy strategy and priority rules, combined with the comprehensive potential field value of the airspace level, to obtain the initial path of the aircraft route includes: Generate a path skeleton through the path nodes in the highway gravitational field, and use the A* algorithm to connect the path nodes to obtain a preliminary path; Starting from the starting point of the aircraft, the initial path of the aircraft route is obtained by iteratively selecting the path node with the lowest comprehensive potential field value in the adjacent three-dimensional grids and optimizing the preliminary path in combination with the pre-set priority rules.

9. The method for low-altitude aircraft route planning in coordination between highway and airspace according to claim 1, characterized in that: The initial path of the aircraft route is adjusted according to the artificial potential field local optimization algorithm and the dynamic conflict detection mechanism to obtain the aircraft route path, including: Based on the principle of gradually reducing the value of the comprehensive potential field, the gradient descent method is used to iteratively adjust the path node positions in the initial path of the aircraft route; Divide the airspace into a space-time grid and predict the future trajectories occupied by aircraft and obstacles; When there are multiple aircraft trajectories in the same space-time grid, and the predicted distance between the aircraft trajectory and the obstacle is less than the static obstacle safety distance, the current path of the aircraft route is determined to be in a conflict state, and the current path of the aircraft route is adjusted according to the pre-set dynamic adjustment strategy to obtain the aircraft route path.

10. The method for low-altitude aircraft route planning in coordination between highway and airspace according to claim 9, characterized in that: The method of iteratively adjusting the path node positions in the initial path of the aircraft route by using the gradient descent method based on the principle of gradually reducing the comprehensive potential field value includes: The node position adjustment formula is used to adjust the path node position in the initial path of the aircraft route, and the node position adjustment formula is: Where P i+1 is the three-dimensional coordinate of the path node at the i+1th iteration, P i is the three-dimensional coordinate of the path node at the i-th iteration; η is the learning rate; is the comprehensive potential field at node P i The gradient vector at .

Images