Flying car formation control method and system, electronic equipment and storage medium

Abstract

The embodiment of the invention provides an aerocar formation control method and system, electronic equipment and a storage medium, and belongs to the technical field of intelligent traffic control. According to the scheme, a plurality of hovercars are linearly formed, the first hovercar in the formation is determined as a head car, the hovercars following the head car in the formation are determined as following cars, and a hovercar formation is obtained; acquiring environment state information and the state quantity of the hovercar; determining a motion model corresponding to the hovercar according to the state quantity; performing path planning according to a preset starting point and an end point to obtain a reference path; based on the reference path, obstacle avoidance planning is carried out through an optimization algorithm according to the environment state information and the motion model, and an obstacle avoidance path is obtained; controlling the hovercar formation to perform dynamic obstacle avoidance according to the obstacle avoidance path; and after avoiding the obstacle, controlling the hovercar formation to return to the reference path. According to the invention, the flight complexity can be reduced, the formation is easy to control and maintain, and the obstacle avoidance safety and the personnel comfort are improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent traffic control technology, and in particular to a flying car formation control method, system, electronic device and storage medium. Background Art

[0002] Low-altitude transportation mainly relies on various low-altitude flight activities, both manned and unmanned, to promote the further development of related industries. Flying cars emerged with this concept, mainly to solve the problem of urban traffic congestion.

[0003] Currently, distributed control is widely used in urban low-altitude transportation. Each flying car makes autonomous obstacle avoidance decisions based on local information, achieving coordination through consensus algorithms. However, in a flying car formation, local optimization of distributed control can lead to redundant paths for the entire formation (i.e., some nodes choose to detour for optimization). Furthermore, as the number of vehicles in the formation increases, communication complexity increases, requiring high-frequency exchange of status information between different nodes. For example, a 10-car formation needs to exchange 1,000 messages per second. When the formation is large, existing 5G networks have high communication latency, making it difficult to react quickly to obstacles. Summary of the Invention

[0004] The main purpose of the embodiments of the present application is to propose a flying car formation control method, system, electronic equipment and storage medium, which aims to solve the problems of dynamic obstacle avoidance, path optimization and multi-machine coordination in formation flight, reduce the complexity of flight, facilitate the control and maintenance of formations, and improve the safety of obstacle avoidance and personnel comfort.

[0005] To achieve the above objectives, one aspect of an embodiment of the present application provides a method for controlling a flying car formation, the method comprising:

[0006] Performing a linear formation on a plurality of flying cars, determining the first flying car in the formation as a lead car, and determining the flying cars following the lead car in the formation as follower cars, thereby obtaining a flying car formation;

[0007] Acquiring environmental status information and a state quantity of the flying car;

[0008] Determining a motion model corresponding to the flying car according to the state quantity;

[0009] Plan the path based on the preset starting point and end point to obtain a reference path;

[0010] Based on the reference path, performing obstacle avoidance planning according to the environmental state information and the motion model through an optimization algorithm to obtain an obstacle avoidance path;

[0011] Controlling the flying car formation to perform dynamic obstacle avoidance according to the obstacle avoidance path;

[0012] After avoiding the obstacle, the flying car formation is controlled to return to the reference path.

[0013] In some embodiments, the state quantities include flight speed, flight path angle, heading angle, bank angle, aerodynamic drag, and engine thrust. Determining the motion model corresponding to the flying car based on the state quantities includes the following steps:

[0014] determining a position state of the flying car in an inertial reference system according to the flight speed, the flight path angle, and the heading angle, and obtaining a position change;

[0015] determining a speed state of the flying car according to the aerodynamic drag, the engine thrust, and the flight path angle, and obtaining a speed change;

[0016] determining an attitude angle state of the flying car according to the flight speed, the flight path angle, the heading angle, and the tilt angle, and obtaining an attitude angle change;

[0017] A motion model corresponding to the flying car is obtained according to the position change, the speed change, and the attitude angle change.

[0018] In some embodiments, the step of performing obstacle avoidance planning based on the reference path and the environment state information and the motion model using an optimization algorithm to obtain an obstacle avoidance path includes the following steps:

[0019] Extract features from environmental state information to obtain environmental features;

[0020] Inputting the environmental features into a long short-term memory neural network to capture the temporal dynamics of the environment and obtain a predicted obstacle trajectory;

[0021] Based on the reference path, performing collision prediction according to the obstacle trajectory and the motion model to obtain a collision prediction result;

[0022] According to the collision prediction result, the reference path is extended by using an octopus tentacle algorithm to obtain a candidate set, wherein the candidate set includes a plurality of candidate paths for obstacle avoidance;

[0023] A path optimization process is performed on the candidate set to obtain an obstacle avoidance path, where the obstacle avoidance path is the optimal candidate path in the candidate set.

[0024] In some embodiments, performing collision prediction based on the reference path, the obstacle trajectory, and the motion model to obtain a collision prediction result includes the following steps:

[0025] Determining whether the obstacle trajectory and the reference path intersect at a next moment;

[0026] When the obstacle trajectory intersects the reference path at the next moment, the flying car with the collision risk is determined according to the motion model to obtain a collision prediction result.

[0027] In some embodiments, performing path optimization on the candidate set to obtain an obstacle avoidance path includes the following steps:

[0028] Performing path optimization processing on the candidate set by a reinforcement learning algorithm to obtain an optimized set, wherein the optimized set includes a plurality of candidate paths that meet basic obstacle avoidance conditions;

[0029] Taking trajectory flight time, path length, path curvature and path mutation as constraints, a multi-objective optimization framework is used to perform weighted processing on each candidate path in the optimization set to obtain a weighted result;

[0030] The candidate path with the best weighted result is determined as the obstacle avoidance path.

[0031] In some embodiments, when the flying cars with collision risks in the collision prediction results include the lead car, controlling the flying car formation to perform dynamic obstacle avoidance according to the obstacle avoidance path includes the following steps:

[0032] determining the leading vehicle motion information according to the obstacle avoidance path and the motion model corresponding to the leading vehicle, and controlling the leading vehicle to perform an obstacle avoidance action according to the leading vehicle motion information;

[0033] The action information of the leading vehicle is sent to the following vehicle through intra-team communication, and the following vehicle is controlled to perform obstacle avoidance action according to the action information of the leading vehicle.

[0034] In some embodiments, performing path optimization processing on the candidate set using a reinforcement learning algorithm to obtain an optimized set includes the following steps:

[0035] A comprehensive reward function is obtained based on positive rewards and negative rewards, wherein the positive rewards include rewards for each action step and rewards for reaching the end point, and the negative rewards are collision penalty rewards;

[0036] Calculating the reward value corresponding to each candidate path in the candidate set by using the comprehensive reward function;

[0037] The candidate set is screened according to the reward value to obtain an optimized set.

[0038] To achieve the above objectives, another aspect of the present application provides a flying car formation control system, the system comprising:

[0039] The first module is configured to form a linear formation of a plurality of flying cars, determine the first flying car in the formation as the lead car, and determine the flying cars following the lead car in the formation as following cars, thereby obtaining a flying car formation;

[0040] The second module is used to obtain environmental status information and the state quantity of the flying car;

[0041] A third module is used to determine a motion model corresponding to the flying car according to the state quantity;

[0042] The fourth module is used to plan a path based on the preset starting point and end point to obtain a reference path;

[0043] A fifth module is configured to perform obstacle avoidance planning based on the reference path using an optimization algorithm according to the environmental state information and the motion model to obtain an obstacle avoidance path;

[0044] A sixth module is configured to control the flying car formation to perform dynamic obstacle avoidance according to the obstacle avoidance path;

[0045] The seventh module controls the flying car formation to return to the reference path after avoiding obstacles.

[0046] To achieve the above-mentioned purpose, another aspect of an embodiment of the present application provides an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned method when executing the computer program.

[0047] To achieve the above-mentioned purpose, another aspect of an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program implements the above-mentioned method when executed by a processor.

[0048] Embodiments of the present application include at least the following beneficial effects: This application provides a flying car formation control method, system, electronic device, and storage medium. This solution forms a linear formation of several flying cars, identifies the first flying car in the formation as the lead car, and identifies the flying cars following the lead car as the following cars, thereby obtaining a flying car formation. The method then obtains environmental state information and flying car state quantities; determines a corresponding motion model for the flying cars based on the state quantities; plans a path based on preset starting and ending points to obtain a reference path; and based on the reference path, performs obstacle avoidance planning using an optimization algorithm based on the environmental state information and the motion model to obtain an obstacle avoidance path. The method controls the flying car formation to perform dynamic obstacle avoidance according to the obstacle avoidance path; and after avoiding obstacles, controls the flying car formation to return to the reference path. This application reduces flight complexity, facilitates control and maintenance of the formation, and improves obstacle avoidance safety and passenger comfort. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a flow chart of a flying car formation control method provided by an embodiment of the present application;

[0050] Figure 2 is a schematic diagram of a flying car formation provided by an embodiment of the present application;

[0051] Figure 3 Schematic diagram of the framework of the obstacle avoidance algorithm provided in the embodiment of the present application;

[0052] Figure 4 is a flow chart of a flying car formation control method provided in another embodiment of the present application;

[0053] Figure 5 is a schematic diagram of static obstacle avoidance provided by an embodiment of the present application;

[0054] Figure 6 is a schematic diagram of dynamic obstacle avoidance provided by an embodiment of the present application;

[0055] Figure 7 This is a schematic diagram of the structure of a flying car formation control system provided by an embodiment of the present application;

[0056] Figure 8 This is a schematic diagram of the hardware structure of the electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application. They are merely examples of devices and methods consistent with some aspects of the embodiments of the present application as detailed in the appended claims.

[0058] It will be understood that the terms "first", "second", etc. used in this application may be used herein to describe various concepts, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0059] The terms "at least one", "plurality", "each", "any", etc. used in this application include "at least one", "two" or more, "plurality" or "each", "any" or "any one", "each" or "any one" in the context of the present invention, and "at least one" or "at least one" includes one, two or more, "plurality" or "any one" includes two or more, "each" or "each one" in the context of the present invention, and "any" or "any one

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0061] Before describing the embodiments of the present application in detail, some related technologies involved in the embodiments of the present application are first described.

[0062] Flying car formations can be mainly divided into three categories: centralized, distributed and hybrid.

[0063] 1. Centralized control:

[0064] Principle: The master node plans the path and distributes instructions, while the other slave nodes only execute instructions.

[0065] Advantages: ① The communication structure is simple, requiring only one-way communication between the master node and the slave node; ② Strong global optimization capability, which can avoid local conflicts.

[0066] Disadvantages: ① The failure of the master node may cause the formation to collapse; ② When the number of formations is large, the computing load of the master node increases dramatically; when there are sudden obstacles (such as birds rushing in), the response delay of the master node is significant.

[0067] 2. Distributed control:

[0068] Principle: Each flying car makes autonomous decisions based on local information (such as neighbor locations and sensor data) and achieves collaboration through a consensus algorithm.

[0069] Advantages: ① Failure of a local node will not affect the overall operation of the formation; ② Each node can quickly avoid obstacles based on distributed model predictive control.

[0070] Disadvantages: ① Local optimality may lead to overall path redundancy of the formation (i.e., some nodes choose to detour for optimization); ② High communication complexity, requiring high-frequency exchange of status information between different nodes.

[0071] 3. Hybrid control:

[0072] Principle: Combining centralized and distributed systems, the upper layer is planned by the central node for the global path, and the lower layer is independently adjusted by each node for details.

[0073] Advantages: ① Balance global optimization and local flexibility, the central node plans the approximate route, and each node achieves autonomous obstacle avoidance; ② Reduces communication load.

[0074] Disadvantages: ① Conflicts may occur between global instructions and local instructions.

[0075] Currently, distributed control is widely used in urban low-altitude transportation. The main disadvantage is the high complexity of communication. For example, a formation of 10 aircraft needs to exchange 1,000 messages per second. When the formation is large, the communication delay is high under the existing 5G network communication, making it difficult to respond quickly when encountering obstacles.

[0076] In view of this, embodiments of the present application provide a flying car formation control method, system, electronic device, and storage medium. This solution forms a linear formation of several flying cars, identifies the first flying car in the formation as the lead car, and identifies the flying cars following the lead car as the following cars, thereby obtaining a flying car formation. Environmental state information and flying car state quantities are obtained; a motion model corresponding to the flying cars is determined based on the state quantities; a path is planned based on preset starting and end points to obtain a reference path; based on the reference path, obstacle avoidance planning is performed using an optimization algorithm based on the environmental state information and the motion model to obtain an obstacle avoidance path; the flying car formation is controlled to perform dynamic obstacle avoidance according to the obstacle avoidance path; and after avoiding obstacles, the flying car formation is controlled to return to the reference path. This application can reduce flight complexity, facilitate control and maintenance of the formation, and improve obstacle avoidance safety and personnel comfort.

[0077] The flying car formation control method provided in the embodiments of the present application relates to the field of intelligent traffic control technology. The flying car formation control method provided in the embodiments of the present application can be applied to a terminal or a server, or can be software running on a terminal or server. In some embodiments, the terminal can be a smartphone, tablet computer, laptop computer, desktop computer, smart speaker, smartwatch, or in-vehicle terminal, etc., but is not limited to these. The server can be configured as an independent physical server, or as a server cluster or distributed system consisting of multiple physical servers. It can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network. The software can be an application that implements the flying car formation control method, etc., but is not limited to the above forms.

[0078] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and the like. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments in which tasks are performed by remote processing devices connected via a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0079] Figure 1 This is an optional flow chart of the flying car formation control method provided in an embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S107.

[0080] In step S101, a plurality of flying cars are linearly formed, the first flying car in the formation is determined as the lead car, and the flying cars following the lead car in the formation are determined as following cars, thereby obtaining a flying car formation.

[0081] Step S102: Acquire environmental status information and the state of the flying car.

[0082] Step S103: determining a motion model corresponding to the flying car according to the state quantity.

[0083] Step S104: performing path planning according to the preset starting point and end point to obtain a reference path.

[0084] Step S105 , based on the reference path, obstacle avoidance planning is performed according to the environment state information and the motion model through an optimization algorithm to obtain an obstacle avoidance path.

[0085] Step S106: Control the flying car formation to perform dynamic obstacle avoidance according to the obstacle avoidance path.

[0086] Step S107: After avoiding the obstacle, control the flying car formation to return to the reference path.

[0087] In this embodiment, the flying cars utilize hybrid control. This involves centralized control for path planning, with the lead vehicle performing rough route planning. Distributed control is employed for obstacle avoidance, with each node (i.e., the flying car) making autonomous decisions when encountering an obstacle. Given the high complexity of commonly used hybrid control architectures, a linear formation similar to an ant colony is employed to reduce control complexity. The first flying car in the formation is designated as the lead vehicle, and the remaining flying cars following the lead vehicle are designated as followers, resulting in a flying car formation.

[0088] For example, Figure 2 As shown, the five flying cars in a linear formation are arranged in a straight line, with the lead car in front and the following cars following at a fixed interval. When planning a path, only one route needs to be planned for the lead car, and the following cars automatically form a formation by maintaining a fixed interval. This simplified navigation and control of the linear formation makes path planning and flight control relatively simple, reducing complex spatial calculations. Furthermore, the flying cars maintain a straight line, reducing flight complexity and making it easier to control and maintain a stable formation.

[0089] Furthermore, linear formations can reduce the risk of collisions. Flying cars fly in a straight line, reducing the likelihood of collisions in space. Furthermore, when encountering obstacles or needing to change flight direction, it's relatively simple for flying cars in a linear formation to maneuver and avoid them, making it easier to adjust their trajectory.

[0090] Through linear formation, flying cars can collect data consistently in the same direction, which is more conducive to the base station's monitoring of the flying car's trajectory, thereby improving data collection efficiency.

[0091] The flying car obtains its own speed, acceleration and other information through the onboard inertial system and communication with GPS as the state variables of the vehicle; and environmental status information such as obstacle position and relative position of the preceding vehicle is obtained through sensor fusion (depth camera + lidar).

[0092] The motion model of the flying car is established according to the state quantity through the dynamic equation, and the state changes of the flying car in the short term in the future are predicted in real time.

[0093] The path planning part is based on GPS information. The starting and ending points are set at the ground station. The lead vehicle integrates the urban terrain distribution to optimize the path and determines the most efficient flight path as the reference path.

[0094] With the help of a depth camera, the environmental status information is captured, and then the captured information is input into a deep learning neural network. Based on the deep learning neural network, the obstacle path is predicted through sensor data fusion. Once the motion model determines that the predicted obstacle path and the reference path of the formation flight have a greater collision risk at the next moment, a new obstacle avoidance path is calculated in real time to avoid static and dynamic obstacles.

[0095] The deviation between the flying car and the planned path point is calculated according to the obstacle avoidance path, and the deviation is input into the PID controller to calculate the control quantity to control the flying car formation to perform dynamic obstacle avoidance.

[0096] After the flying car formation avoids the obstacle according to the obstacle avoidance path, the flying cars are controlled to return to the reference path and restore to the original linear formation.

[0097] In some embodiments, step S103 may include but is not limited to steps S201 to S204.

[0098] Step S201 : determining the position state of the flying car in an inertial reference system according to the flight speed, flight path angle, and heading angle to obtain a position change.

[0099] Step S202: determining the speed state of the flying car based on aerodynamic drag, engine thrust, and flight path angle to obtain a speed change.

[0100] Step S203: determining the attitude angle state of the flying car according to the flight speed, flight path angle, heading angle, and bank angle, and obtaining an attitude angle change.

[0101] Step S204: Obtain a motion model corresponding to the flying car based on the position change, speed change, and attitude angle change.

[0102] In this embodiment, the state quantity can be divided into position state, velocity state and attitude angle state.

[0103] The position state is the three-dimensional space coordinate x of the flying car in the inertial system i ,y i , z i The current position state can be obtained in real time through GPS positioning. The future position state of the flying car is determined in the inertial reference system based on the flight speed, flight path angle and heading angle, and the position change is obtained.

[0104] The speed state is the flight speed, the current speed state V i The airspeed is measured by an airspeed meter, while the dynamic parameters of the flying car, such as aerodynamic drag and engine thrust, require dynamic models and sensor measurements. The future speed state of the flying car is determined based on aerodynamic drag, engine thrust, and flight path angle, resulting in the speed change.

[0105] The attitude angle state includes γ i (flight path angle, i.e. the angle between the velocity vector and the horizontal plane), χ i (heading angle, i.e. the angle between the projection of the velocity vector on the horizontal plane and the reference direction) and φ i The pitch angle is the rotation angle of the flying car around its longitudinal axis. The pitch angle can be directly obtained from IMU (Inertial Measurement Unit) data, while the flight path angle and heading angle can be calculated using a formula. The future attitude angle state of the flying car is determined based on the flight speed, flight path angle, heading angle, and pitch angle, and the attitude angle change is obtained.

[0106] The motion model corresponding to the flying car is obtained based on the position change, speed change, and attitude angle change. The expression of the motion model is as follows:

[0107]

[0108] Where i = 1, 2, 3, ..., represents the i-th car in the flying car formation, x i 、y i 、z i V represents the change in the three-dimensional space coordinates of the i-th flying car in the inertial system, i , γ i , χ i Represent the changes in flight speed, flight path angle and heading angle respectively, D i Expressed as the aerodynamic drag of the flying car, T i is the engine thrust, n i is the load factor, φ i is the tilt angle of the flying car, m is the empty mass of the flying car, and g is the acceleration due to gravity.

[0109] In some embodiments, step S105 may include but is not limited to steps S301 to S305.

[0110] Step S301: extract features from environmental status information to obtain environmental features.

[0111] Step S302: Input the environmental features into the long short-term memory neural network to capture the temporal dynamics of the environment and obtain the predicted obstacle trajectory.

[0112] Step S303 : performing collision prediction based on the reference path, obstacle trajectory and motion model to obtain a collision prediction result.

[0113] Step S304 : Based on the collision prediction result, the reference path is expanded by using the octopus tentacle algorithm to obtain a candidate set, which includes several candidate paths for obstacle avoidance.

[0114] Step S305 , performing path optimization processing on the candidate set to obtain an obstacle avoidance path, where the obstacle avoidance path is the optimal candidate path in the candidate set.

[0115] Reference Figure 3 In this embodiment, the flying car normalizes the captured environmental state information, and then inputs the normalized information into the CNN network for preprocessing. The network is used to extract key features such as semantics and boundaries in the captured environment to obtain environmental features.

[0116] These extracted key features are sent to the long short-term memory neural network (LSTM) layer, which is responsible for storing and processing the historical state information of the flying car to capture the temporal dynamics of the environment and predict obstacle trajectories.

[0117] The outputs of the LSTM layer and the CNN network are combined as training data and fed into the main program to implement collision prediction. The trained collision prediction model uses the reference path, obstacle trajectory, and motion model to predict collisions. It determines whether there is a collision risk between the flying car formation and the obstacle, and generates a collision prediction result.

[0118] When the collision prediction result indicates a high collision risk, the reference path is expanded using the Octopus algorithm. The Octopus algorithm is an optimization algorithm that simulates the tentacles of an octopus. Based on a set of randomly generated initial solutions that meet basic obstacle avoidance conditions, a series of domain solutions are expanded. Different candidate paths are obtained by combining different domain solutions. All the obtained candidate paths are summarized to form a candidate set.

[0119] In order to ensure the smoothness of the path and reduce sharp turns or unnecessary complex paths, the candidate set is optimized, and a candidate path that meets the comprehensive requirements of efficient, safe and comfortable flight of the formation is found from all candidate paths as the obstacle avoidance path.

[0120] In some embodiments, step S303 may include but is not limited to steps S401 to S402.

[0121] Step S401: determine whether the obstacle trajectory and the reference path intersect at the next moment.

[0122] Step S402: When the obstacle trajectory intersects the reference path at the next moment, a flying car with a collision risk is determined based on the motion model to obtain a collision prediction result.

[0123] In this embodiment, the collision prediction determines whether there is a collision risk by detecting whether the obstacle trajectory predicted by the LSTM network intersects with the reference path at the next moment.

[0124] Specifically, the obstacle trajectory and the reference path are unified into the same coordinate system and their data are aligned by timestamp. For each prediction moment, the distance between the obstacle and the reference path is calculated. If the distance is less than a preset safety threshold, it is considered that the obstacle trajectory and the reference path have a path intersection at that prediction moment.

[0125] Furthermore, when the obstacle trajectory intersects the reference path at the next moment, the position state of the flying car at the future moment is predicted based on the motion model of each flying car, the flying car with collision risk is determined, and the collision prediction result is obtained to avoid static obstacles and dynamic obstacles.

[0126] For example, if the obstacle is a static obstacle, when the obstacle trajectory intersects with the reference path, it can be determined based on the motion model that all flying cars in the formation are at risk of collision.

[0127] If the obstacle is dynamic, then when the obstacle trajectory intersects the reference path, the motion model can be used to determine which flying cars in the formation are at risk of collision with the obstacle, so that the flying cars at risk can be accurately controlled to avoid the obstacle, while the remaining flying cars maintain the formation and continue to fly along the reference path.

[0128] In some embodiments, step S305 may include but is not limited to steps S501 to S503.

[0129] Step S501 : Optimize the candidate set by using a reinforcement learning algorithm to obtain an optimized set. The optimized set includes several candidate paths that meet basic obstacle avoidance conditions.

[0130] In step S502 , a multi-objective optimization framework is used to perform weighted processing on each candidate path in the optimization set, taking trajectory flight time, path length, path curvature, and path mutation as constraints, to obtain a weighted result.

[0131] Step S503: Determine the candidate path with the best weighted result as the obstacle avoidance path.

[0132] In this embodiment, a reinforcement learning algorithm is used to optimize paths. The candidate paths in the candidate set are screened, and a set of candidate paths that meet basic obstacle avoidance requirements is identified. These requirements include the flying car flying toward its destination, the flight time being less than or equal to a preset time threshold, and no collisions occurring during flight. The selected candidate paths that meet these basic obstacle avoidance requirements are then aggregated to form an optimized set.

[0133] In some embodiments, step S501 may include but is not limited to steps S601 to S603.

[0134] Step S601: Obtain a comprehensive reward function based on positive rewards and negative rewards, wherein the positive rewards include rewards for each action and rewards for reaching the end point, and the negative rewards are collision penalty rewards.

[0135] Step S602: Calculate the reward value corresponding to each candidate path in the candidate set using a comprehensive reward function.

[0136] Step S603: Screen the candidate set according to the reward value to obtain the optimized set.

[0137] Specifically, this embodiment proposes a comprehensive reward r that includes two sub-goals: positive reward and negative reward. i , the positive reward at time t includes the reward r for each action of the flying car t act and the reward for reaching the end point and achieving the goal t success , negative rewards are the punishment for collision behavior r t col Specific reward and punishment formulas include:

[0138] ① The calculation formula for each action reward:

[0139] r t act =e -PdP (2);

[0140] Among them, d refers to the distance from the flying car at a certain moment to the end point S d (x d ,y d ,zd)

[0141] When the flying car always flies towards the destination, the agent is given a gradually increasing reward; on the contrary, if the flying car moves away from the destination, the agent is given a gradually decreasing reward to regulate the flying car's actions and achieve the expected goal.

[0142] ② Reward calculation formula for reaching the end point:

[0143]

[0144] Among them, T is the total time required for the flying car to reach the destination. When this time does not exceed the specified threshold T L When the flying car reaches the destination within the specified time, the agent will receive a reward of score; otherwise, the agent will not receive any reward.

[0145] ③ Collision penalty calculation formula:

[0146]

[0147] If the flying car collides with a dynamic obstacle on the way to the end point, it will be penalized accordingly. That is, when the flying car hits an obstacle, it will receive a negative reward of -score to encourage the agent to avoid such behavior.

[0148] ④Calculation formula of comprehensive reward function:

[0149] r t =w1*r t act +w2*r t col +w3*r t success (5);

[0150] Among them, w1, w2, and w3 are the weight factors of each sub-goal reward.

[0151] The reward value of each candidate path is calculated according to formula (5), and multiple candidate paths with higher scores are screened out according to the reward value to obtain the optimization set.

[0152] Furthermore, after the candidate paths generated by the Octopus algorithm are initially screened by the reinforcement learning algorithm to obtain the optimized set, the optimized set is further screened using the multi-objective framework. The trajectory flight time, path length, path curvature and path mutation are used as constraints. The weights of the above multiple parameters are reasonably allocated to weight the candidate paths in the optimized set to obtain the weighted results.

[0153] Specifically, the trajectory flight time and path length participate in the reward value calculation of the reinforcement learning in the aforementioned step S501, so the reward value calculated by the comprehensive reward function in the reinforcement learning can be used to represent the trajectory flight time and path length.

[0154] Path curvature is the ratio of the tangent angle to the arc length at a point on the path. Defined by differentiation, it indicates the degree of curvature of the path at that point and reflects the degree to which the curve deviates from a straight line. A greater curvature indicates a more curved curve.

[0155] For example, the path curvature r t k It can be calculated based on the change in the angle between two adjacent points on the curve and the arc length between the two adjacent points. The calculation formula is:

[0156]

[0157] Among them, dθ is the change in the angle between two adjacent points, and ds is the arc length between two adjacent points.

[0158] Path mutation usually refers to the sudden change of trajectory direction or speed of a flying car during flight due to external interference (such as obstacle avoidance), control command switching or path planning algorithm limitations.

[0159] Exemplarily, the causes of path mutation include:

[0160] (1) Model discontinuity in kinematics: In path planning, if the generated trajectory does not meet the continuity conditions in position, velocity or acceleration (for example, the polynomial trajectory is of insufficient order), it may lead to sudden changes in the path.

[0161] (2) Path switching caused by obstacle avoidance (main): When the UAV detects an obstacle, it may generate a new path through algorithms (such as RRT*, artificial potential field method), resulting in a mutation of the original path.

[0162] (3) Constraint conflict in path optimization: When optimizing a trajectory (e.g., minimizing energy or time), if the constraints (e.g., maximum speed, curvature) are suddenly violated, it may lead to sudden changes in path replanning.

[0163] This sudden change may manifest as discontinuous changes in heading angle, sudden acceleration, or discontinuity in path curvature.

[0164] Specifically, a path mutation can be represented by a change in heading angle, velocity, acceleration, or curvature that exceeds a set threshold within a very short Δt. The above mutation can be calculated based on the trajectory parameters.

[0165] (1) Position mutation: It refers to the sudden change of the position vector direction or distance between adjacent path points.

[0166] ① Direction change: Calculate the rate of change of heading angle between adjacent path segments. The calculation formula is as follows:

[0167] Δθ=θ i+1 -θ i (7);

[0168] Among them, θ i is the heading angle of the i-th path. If |Δθ| exceeds a threshold (e.g., 30°), it is considered a sudden change in direction.

[0169] ② Distance mutation: Calculate the displacement difference change rate of adjacent path points. The calculation formula is as follows:

[0170] Δd=p i+1 -p i (8);

[0171] Among them, P i is the displacement of the i-th path. If Δd is significantly larger than the average step length, it is considered a distance mutation.

[0172] (2) Acceleration jerk: It is defined as the discontinuity of the derivative of acceleration. The calculation formula is as follows:

[0173]

[0174] Among them, a t Indicates the acceleration at time t. If the Jerk exceeds the threshold (such as 10m / s 2 ), it is considered as a sudden acceleration change.

[0175] (3) Sudden change in curvature: This refers to a sudden change in the curvature of the trajectory (such as a sharp turn when avoiding an obstacle). The calculation formula is as follows:

[0176] Δκ=|κ i+1 -κ i | (10);

[0177] Among them, K i represents the curvature of the i-th path, If the threshold is exceeded (such as 0.5m -1 ), it is considered as a sudden change in curvature.

[0178] The mutations obtained in different dimensions are normalized. The smaller the results of the normalized position mutation, acceleration mutation, and curvature mutation, the lower the degree of path mutation and the more comfortable the path. Therefore, each specific mutation result is inversely proportional to the score of the path mutation, that is, the smaller the mutation, the higher the score.

[0179] In summary, the reward value, path curvature and path mutation are weighted and summed to obtain the weighted result, the weighted result r′ t The calculation formula is:

[0180] r′ t =w4*r t +w5*r t k +w6*r t mut (11);

[0181] Among them, w4, w5, and w6 are the weight factors of reward value, path curvature, and path mutation, respectively.

[0182] It is understandable that the various weight factors can be set as needed based on the comprehensive requirements of efficient, safe and comfortable flight of the formation.

[0183] The candidate paths are sorted according to the weighted results, and the candidate path with the highest weighted result score is determined as the optimal candidate path to obtain the obstacle avoidance path.

[0184] In some embodiments, step S106 may include but is not limited to steps S701 to S702.

[0185] Step S701 : determining the action information of the leading vehicle according to the obstacle avoidance path and the motion model corresponding to the leading vehicle, and controlling the leading vehicle to perform an obstacle avoidance action according to the action information of the leading vehicle.

[0186] In step S702, the action information of the leading vehicle is sent to the following vehicle through intra-team communication, and the following vehicle is controlled to perform obstacle avoidance action according to the action information of the leading vehicle.

[0187] In this embodiment, when the flying cars with collision risks in the collision prediction results include the lead car, the lead car determines the lead car action information based on the obstacle avoidance strategy according to the obstacle avoidance path and the corresponding motion model, and performs obstacle avoidance action.

[0188] The following vehicle obtains the action information of the leading vehicle in real time through intra-team communication and performs obstacle avoidance actions based on its own position.

[0189] The following describes and explains the solution of the embodiment of the present invention in detail with reference to specific application examples.

[0190] Reference Figure 3 , Figure 3 Another embodiment of the present application provides a flying car formation control method, which is applied to a flying car. The flying car includes three modules: a trajectory following module, an obstacle avoidance mechanism module, and an optimization algorithm module.

[0191] The trajectory following module is used to establish the motion model of the flying car, determine the state quantity of the motion model, and ensure that the flying car flies according to the predetermined path.

[0192] The obstacle avoidance mechanism module is based on deep learning neural networks, and collision prediction and dynamic path adjustment are performed through sensor data fusion to achieve real-time obstacle avoidance.

[0193] By optimizing the path based on reinforcement learning through the optimization algorithm module, the flying car improves flight efficiency and reduces energy consumption through interaction with the real environment.

[0194] Specifically, the trajectory following module establishes a motion model as shown in formula (1), using all the state parameters in the model. GPS positioning data is used to correct the position state (xi, yi, zi), and IMU data (acceleration, angular velocity) is used to calculate the speed and attitude change rate (V i ,γ i ,χ i ).

[0195] The trajectory following module itself includes a path planning algorithm and a control variable input calculation section. The path planning section uses GPS information to set the starting and ending points at the ground station, integrating the urban terrain distribution to optimize the path and determine the most efficient flight path. The control variable input calculation section first calculates the GPS coordinates of the flying car, obtains the deviation from the planned path points, and inputs this into the PID controller for calculation and control, ensuring that the vehicles in the formation follow the planned path.

[0196] The obstacle avoidance module is based on a deep learning neural network and uses sensor data fusion to perform obstacle avoidance control. The sensor data fusion step includes designing the flying car's state space and action space.

[0197] (1) Design of flying car state space:

[0198] The flying car obtains its own speed, acceleration and other information through onboard inertial navigation and communication with GPS as its state variables; and environmental information such as obstacle position and relative position of the preceding vehicle is obtained through sensor fusion (depth camera + lidar) and used in the obstacle avoidance mechanism module, but does not directly affect the state variables of the motion model.

[0199] (2) Design of flying car action space:

[0200] Flying cars interact with their environment by performing specific actions, causing changes in the environmental state. For autonomous obstacle avoidance and navigation tasks, the use of a continuous action space is a key design consideration, enabling more precise and fluid motion control. In this invention, the action space is divided into three dimensions, and the action of the flying car at time t is defined as:

[0201] a t =[v xt ;v yt ;v zt ](12);

[0202] Among them, v xt 、v yt and v zt Represents the linear speed of a flying car on the x, y, and z axes.

[0203] like Figure 3 As shown in the figure, at time t, the flying car uses the depth camera to capture some environmental status information and transmits the captured environmental status information to the LSTM network. The LSTM network predicts the obstacle path to achieve collision prediction. Once the obstacle path predicted by the LSTM network intersects with the flight path of the formation at the next moment, there is a greater risk of collision, and the system will immediately calculate a new flight path.

[0204] In this embodiment, the Octopus algorithm is used as an obstacle avoidance algorithm to participate in the path optimization design of the lead vehicle, and different obstacle avoidance paths are obtained, so that the optimal path can be found as the obstacle avoidance path based on the constraints such as path length and path curvature, so as to achieve efficient and safe obstacle avoidance of the formation.

[0205] The optimization algorithm module uses reinforcement learning to optimize the path, improve flight efficiency and reduce energy consumption. This includes the design of a reward function and the use of a multi-objective optimization framework.

[0206] The comprehensive reward function designed by formula (5) performs weighted calculations for each sub-goal reward, ensuring that the optimized path can ensure that the flying car can not only avoid obstacles autonomously during autonomous flight, but also reach the destination quickly, thereby achieving efficient and safe navigation.

[0207] The multi-objective optimization framework weights the multiple paths generated by the Octopus optimization algorithm using parameters such as flight time, path curvature, and path mutations to ensure that the optimized path meets the comprehensive requirements of efficient, safe, and comfortable flight for the formation. Furthermore, due to airspace restrictions imposed by the air bridge architecture over cities, a tolerance for the altitude z of the flying car formation is required to ensure that the formation can navigate and avoid both static and dynamic obstacles within the required flight path altitude. By rationally weighting multiple objective parameters, the optimal obstacle avoidance path is achieved, ensuring safety, efficiency, and crew comfort.

[0208] It should be noted that, according to the type of obstacle, obstacle avoidance can be divided into following vehicle obstacle avoidance and emergency obstacle avoidance.

[0209] Vehicle-following obstacle avoidance allows a flying car formation to proactively plan and execute avoidance maneuvers when encountering static obstacles in low-altitude urban environments. The following vehicle, through intra-fleet communication, receives real-time information about the lead vehicle's movements and adjusts its flight trajectory based on its own position to effectively avoid static obstacles, ensuring overall safety and coordination within the formation.

[0210] For example, Figure 5 As shown, when the flying car formation faces a bridge, the leading car plans the obstacle avoidance path, and the following cars automatically follow the leading car along the obstacle avoidance path to form a formation. After bypassing the obstacle, the flying car formation resumes the linear formation and continues to fly along the reference path.

[0211] In case of emergency obstacle avoidance, when a flying car formation encounters dynamic obstacles such as birds and falling objects while flying along the route, the depth camera first captures the obstacle information and transmits the information to the long short-term memory network (LSTM) for situational awareness. LSTM analyzes the movement trajectory of the obstacle and predicts its position at the next moment. Figure 6 As shown in the figure, the LSTM network predicts the next moment's position based on the target's previous moment position and current position in the environmental state information and the target's flight direction.

[0212] Once a potential collision with an obstacle is predicted, the deep reinforcement learning obstacle avoidance algorithm is triggered, executing an avoidance maneuver to ensure the flying car can promptly circumvent dynamic obstacles, effectively improving flight safety. In the event of an unexpected obstacle, each flying car at each node can autonomously decide to avoid the obstacle. Only the flying car whose flight path intersects the predicted obstacle path at the next moment will execute the avoidance maneuver. The remaining flying cars in the formation will continue along the reference path. After the flying car that has executed the obstacle avoidance maneuver completes the avoidance, the deviation from the reference path is input into the PID controller for calculation and control, allowing the autonomously avoiding flying car to return to the flying car formation and continue along the planned path.

[0213] In summary, this application uses the LSTM hybrid model to predict obstacle trajectories and combines it with the Octopus algorithm to dynamically adjust the path. The response time can be compressed to less than 0.5 seconds (traditional PID control requires 0.8-1.2 seconds), meeting the acceleration and deceleration requirements of 3m / s2 at low altitude in cities and achieving real-time response.

[0214] At the same time, by using linear formations similar to those of ant colonies, the space occupied by the flying car formation can be greatly reduced (the formation path is basically a single line, while V-shaped, trapezoidal, and circular formations use multiple lines), reducing the possibility of encountering dynamic obstacles. It also supports collaborative obstacle avoidance decisions between the lead vehicle and the following vehicles. In scenarios with obstacle density greater than 5 / km², the dynamic obstacle avoidance success rate is ≥ 94.7%, significantly improving the obstacle avoidance success rate.

[0215] Finally, by combining the multi-objective optimization framework of the Octopus algorithm, unnecessary complex paths were reduced, and the comprehensive energy consumption was reduced by 16.5% compared with traditional PID control.

[0216] Reference Figure 7 The present application also provides a flying car formation control system that can implement the above-mentioned flying car formation control method. The system includes:

[0217] The first module is used to form a linear formation of several flying cars, determine the first flying car in the formation as the lead car, and determine the flying cars following the lead car in the formation as following cars, thereby obtaining a flying car formation.

[0218] The second module is used to obtain environmental status information and the state quantity of the flying car.

[0219] The third module is used to determine the corresponding motion model of the flying car based on the state quantity.

[0220] The fourth module is used to plan a path based on the preset starting point and end point to obtain a reference path.

[0221] The fifth module is used to perform obstacle avoidance planning based on the reference path, using an optimization algorithm according to the environmental state information and the motion model to obtain an obstacle avoidance path.

[0222] The sixth module is used to control the flying car formation to perform dynamic obstacle avoidance according to the obstacle avoidance path.

[0223] The seventh module controls the flying car formation to return to the reference path after avoiding obstacles.

[0224] It can be understood that the contents of the above method embodiments are all applicable to the present system embodiments, the functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0225] The present application also provides an electronic device comprising a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned flying car formation control method. The electronic device can be any smart terminal, including a tablet computer and an in-vehicle computer.

[0226] It can be understood that the contents of the above method embodiments are applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0227] Reference Figure 8 , Figure 8 The hardware structure of an electronic device according to another embodiment is shown. The electronic device includes:

[0228] The processor 901 can be implemented using a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application.

[0229] Memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). Memory 902 can store an operating system and other application programs. When implementing the technical solutions provided in the embodiments of this specification through software or firmware, the relevant program code is stored in memory 902 and is called by processor 901 to execute the flying car formation control method of the embodiments of this application.

[0230] The input / output interface 903 is used to implement information input and output.

[0231] The communication interface 904 is used to realize communication interaction between this device and other devices. Communication can be realized through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0232] The bus 905 transmits information between various components of the device (eg, the processor 901 , the memory 902 , the input / output interface 903 , and the communication interface 904 ).

[0233] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0234] An embodiment of the present application further provides a computer-readable storage medium storing a computer program, which implements the above-mentioned flying car formation control method when executed by a processor.

[0235] It can be understood that the contents of the above method embodiments are all applicable to the present storage medium embodiment, the functions specifically implemented by the present storage medium embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0236] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0237] The embodiments described in the embodiments of this application are intended to more clearly illustrate the technical solutions of the embodiments of this application and do not constitute a limitation on the technical solutions provided by the embodiments of this application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0238] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0239] The system embodiment described above is merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0240] Those skilled in the art will appreciate that all or some of the steps in the methods, systems, and functional modules / units in the devices disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0241] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0242] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and essence of the present invention should be within the scope of the present invention.

Claims

1. A flying car formation control method, characterized in that: The method comprises the following steps: Performing a linear formation on a plurality of flying cars, determining the first flying car in the formation as a lead car, and determining the flying cars following the lead car in the formation as follower cars, thereby obtaining a flying car formation; Acquiring environmental status information and a state quantity of the flying car; Determining a motion model corresponding to the flying car according to the state quantity; Plan the path based on the preset starting point and end point to obtain a reference path; Based on the reference path, performing obstacle avoidance planning according to the environmental state information and the motion model through an optimization algorithm to obtain an obstacle avoidance path; Controlling the flying car formation to perform dynamic obstacle avoidance according to the obstacle avoidance path; After avoiding the obstacle, the flying car formation is controlled to return to the reference path.

2. The method according to claim 1, characterized in that The state quantities include flight speed, flight path angle, heading angle, bank angle, aerodynamic drag, and engine thrust. Determining the motion model corresponding to the flying car based on the state quantities includes the following steps: determining a position state of the flying car in an inertial reference system according to the flight speed, the flight path angle, and the heading angle, and obtaining a position change; determining a speed state of the flying car according to the aerodynamic drag, the engine thrust, and the flight path angle, and obtaining a speed change; determining an attitude angle state of the flying car according to the flight speed, the flight path angle, the heading angle, and the tilt angle, and obtaining an attitude angle change; A motion model corresponding to the flying car is obtained according to the position change, the speed change, and the attitude angle change.

3. The method according to claim 1, characterized in that The method of performing obstacle avoidance planning based on the reference path and the environment state information and the motion model by using an optimization algorithm to obtain an obstacle avoidance path includes the following steps: Extract features from environmental state information to obtain environmental features; Inputting the environmental features into a long short-term memory neural network to capture the temporal dynamics of the environment and obtain a predicted obstacle trajectory; Based on the reference path, performing collision prediction according to the obstacle trajectory and the motion model to obtain a collision prediction result; According to the collision prediction result, the reference path is extended by using an octopus tentacle algorithm to obtain a candidate set, wherein the candidate set includes a plurality of candidate paths for obstacle avoidance; A path optimization process is performed on the candidate set to obtain an obstacle avoidance path, where the obstacle avoidance path is the optimal candidate path in the candidate set.

4. The method according to claim 3, characterized in that The method of performing collision prediction based on the reference path, according to the obstacle trajectory and the motion model, and obtaining a collision prediction result includes the following steps: Determining whether the obstacle trajectory and the reference path intersect at a next moment; When the obstacle trajectory intersects the reference path at the next moment, the flying car with the collision risk is determined according to the motion model to obtain a collision prediction result.

5. The method according to claim 3, characterized in that The performing path optimization processing on the candidate set to obtain an obstacle avoidance path includes the following steps: Performing path optimization processing on the candidate set by a reinforcement learning algorithm to obtain an optimized set, wherein the optimized set includes a plurality of candidate paths that meet basic obstacle avoidance conditions; Taking trajectory flight time, path length, path curvature and path mutation as constraints, a multi-objective optimization framework is used to perform weighted processing on each candidate path in the optimization set to obtain a weighted result; The candidate path with the best weighted result is determined as the obstacle avoidance path.

6. The method according to claim 4, characterized in that When the flying cars with collision risks in the collision prediction results include the lead car, controlling the flying car formation to perform dynamic obstacle avoidance according to the obstacle avoidance path includes the following steps: determining the leading vehicle motion information according to the obstacle avoidance path and the motion model corresponding to the leading vehicle, and controlling the leading vehicle to perform an obstacle avoidance action according to the leading vehicle motion information; The action information of the leading vehicle is sent to the following vehicle through intra-team communication, and the following vehicle is controlled to perform obstacle avoidance action according to the action information of the leading vehicle.

7. The method according to claim 5, characterized in that The path optimization process is performed on the candidate set by using a reinforcement learning algorithm to obtain an optimized set, comprising the following steps: A comprehensive reward function is obtained based on positive rewards and negative rewards, wherein the positive rewards include rewards for each action step and rewards for reaching the end point, and the negative rewards are collision penalty rewards; Calculating the reward value corresponding to each candidate path in the candidate set by using the comprehensive reward function; The candidate set is screened according to the reward value to obtain an optimized set.

8. A flying car formation control system, characterized in that: The system comprises: The first module is configured to form a linear formation of a plurality of flying cars, determine the first flying car in the formation as the lead car, and determine the flying cars following the lead car in the formation as following cars, thereby obtaining a flying car formation; The second module is used to obtain environmental status information and the state quantity of the flying car; A third module is used to determine a motion model corresponding to the flying car according to the state quantity; The fourth module is used to plan a path based on the preset starting point and end point to obtain a reference path; A fifth module is configured to perform obstacle avoidance planning based on the reference path using an optimization algorithm according to the environmental state information and the motion model to obtain an obstacle avoidance path; A sixth module is configured to control the flying car formation to perform dynamic obstacle avoidance according to the obstacle avoidance path; The seventh module controls the flying car formation to return to the reference path after avoiding obstacles.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images