Aircraft approach landing assistance method, apparatus, device, and medium

Abstract

The application provides an aircraft approach and landing auxiliary method, device, equipment and medium, the method comprises the following steps: in the process of multiple aircraft approach and landing, obtain the aircraft radar point cloud data collected by the common sense base station; the common sense base station is arranged at the top of a multi-layer rotating vertical take-off and landing field, and the multi-layer rotating vertical take-off and landing field comprises multiple overhanging take-off and landing platforms; input the aircraft radar point cloud data into a pre-trained strategy prediction model to obtain the common sense base station execution strategy output by the strategy prediction model; according to the common sense base station execution strategy, control the common sense base station to communicate with multiple aircrafts, and control the common sense base station to adjust multiple overhanging take-off and landing platforms. In this way, by adopting a multi-layer rotating vertical take-off and landing field, the floor area is reduced, the aircraft receiving capacity is improved, and the overhanging take-off and landing platforms can be adjusted by the common sense base station, so that the take-off and landing platforms can cooperate with the attitude of the aircraft during approach and landing, and the time of the aircraft hovering in the airspace is reduced.

Description

Technical Field

[0001] This application relates to the field of air navigation technology, and more specifically, to an aircraft approach and landing assistance method, device, equipment, and medium. Background Technology

[0002] With the development of smart city construction and the low-altitude economy, Urban Air Mobility (UAM), as an important component of the future integrated transportation system, is becoming a key area of ​​research and development for governments, research institutions, and industries worldwide. UAM aims to achieve three-dimensional transportation within cities and between city clusters through low-altitude aircraft, thereby effectively alleviating ground traffic congestion, reducing carbon emissions, and improving travel efficiency.

[0003] Aircraft approaching, queuing, and landing in urban low-altitude airspace rely on ground-based vertical takeoff and landing (VTOL) facilities. VTOL facilities often occupy large areas and have a limited number of takeoff and landing platforms, resulting in a limited number of aircraft that can be accommodated at any given time. During approach, aircraft often need to hover at low altitudes and spend considerable time adjusting their attitude to ensure alignment with the platforms, significantly impacting the safety and efficiency of approach and landing. Summary of the Invention

[0004] In view of this, this application provides an aircraft approach and landing assistance method, device, equipment and medium. By adopting a multi-layer rotating vertical take-off and landing field, the footprint is reduced and the aircraft's carrying capacity is increased. Furthermore, multiple externally suspended take-off and landing platforms can be adjusted through a sensing base station, so that the take-off and landing platforms can be coordinated with the aircraft's attitude during approach and landing, thereby improving the safety and efficiency of aircraft approach and landing.

[0005] Specifically, this application is implemented through the following technical solution: According to a first aspect of this application, an approach and landing assistance method for an aircraft is provided, the method comprising: During the approach and landing of multiple aircraft, radar point cloud data of the aircraft collected by the sensing base station is acquired; the sensing base station is set at the top of the multi-layer rotating vertical take-off and landing field, which includes multiple externally suspended take-off and landing platforms. The radar point cloud data of the aircraft is input into a pre-trained strategy prediction model to obtain the sensor base station execution strategy output by the strategy prediction model. The sensor base station execution strategy is used to instruct the sensor base station to perform actions in the process of providing communication services and landing alignment services to the aircraft. According to the execution strategy of the sensing base station, the sensing base station is controlled to communicate with the multiple aircraft, and the sensing base station is controlled to adjust the multiple externally suspended take-off and landing platforms.

[0006] In one optional implementation, the policy prediction model includes a radar point cloud feature extraction unit and a policy generation unit; The step of inputting the aircraft radar point cloud data into a pre-trained policy prediction model to obtain the sensing base station execution policy output by the policy prediction model includes: The radar point cloud feature extraction unit extracts features from the aircraft radar point cloud data to obtain a radar point cloud feature vector. The aircraft radar point cloud data includes the aircraft position, the aircraft azimuth angle, the aircraft pitch angle, the relative distance between the sensing base station and the aircraft, the relative speed between the sensing base station and the aircraft, and the radar signal-to-noise ratio of the aircraft. The strategy generation unit performs strategy reasoning based on the radar point cloud feature vector to obtain the execution strategy of the sensing base station; the execution strategy of the sensing base station includes the communication power allocation ratio between the sensing base station and each aircraft, the horizontal rotation control amount of each externally suspended take-off and landing platform, and the pitch control amount of each externally suspended take-off and landing platform.

[0007] In one optional implementation, the policy prediction model is obtained by training a reinforcement learning network: A state space representing the environmental information of the aircraft's approach and landing is constructed. Based on the state space, training samples are generated, including radar point cloud data of the test aircraft collected by the test sensing base station. The training samples are input into the reinforcement learning network to obtain the test sensory base station execution strategy output by the reinforcement learning network. When executing the test sensing base station execution strategy, determine the platform selection cost corresponding to each test aircraft choosing to land on an externally suspended take-off and landing platform, and determine the reward value corresponding to the test sensing base station execution strategy; Minimizing the platform selection cost is determined as the optimization objective. The reinforcement learning network is iteratively trained until the network meets the cutoff condition based on the reward value. The reinforcement learning network that meets the cutoff condition is then determined as the trained policy prediction model.

[0008] In one optional implementation, the platform selection cost corresponding to landing on an externally suspended takeoff and landing platform for any of the test aircraft is determined through the following steps: Based on the position of the test aircraft and the position of the externally suspended take-off and landing platform selected by the test aircraft for landing, the distance error between the test aircraft and the corresponding externally suspended take-off and landing platform is determined; Based on the azimuth angle of the test aircraft and the corresponding horizontal rotation angle of the externally suspended take-off and landing platform, the horizontal alignment error between the test aircraft and the corresponding externally suspended take-off and landing platform is determined. Based on the pitch angle of the test aircraft and the corresponding pitch angle of the externally suspended take-off and landing platform, determine the pitch direction alignment error between the test aircraft and the corresponding externally suspended take-off and landing platform; Based on the communication bandwidth between the test aircraft and the corresponding external take-off and landing platform and the signal-to-noise ratio of the communication link of the test aircraft, the communication channel capacity between the test aircraft and the corresponding external take-off and landing platform is determined. Based on the distance error, horizontal alignment error, pitch alignment error, and communication channel capacity between the test aircraft and the corresponding externally suspended take-off and landing platform, the platform selection cost corresponding to the test aircraft choosing the externally suspended take-off and landing platform for landing is determined.

[0009] In one optional implementation, the reinforcement learning network is determined to meet the cutoff condition through the following steps: Based on the communication channel capacity, distance error, horizontal alignment error, pitch alignment error, horizontal rotation control amount and pitch control amount of each externally suspended take-off and landing platform between the test aircraft and the corresponding externally suspended take-off and landing platform, the reward value corresponding to the execution strategy of the test sensing base station is determined. When the reward value tends to stabilize, the reinforcement learning network is determined to meet the cutoff condition.

[0010] In one optional implementation, the policy generation unit includes a policy sub-network and a value function sub-network. The policy sub-network is used to output the execution policy of the test sensing base station, and the value function sub-network is used to evaluate the value of the execution policy of the test sensing base station and generate a reward value corresponding to the execution policy of the test sensing base station. Each training round includes multiple consecutive execution steps, and the corresponding execution policy and reward value of the test sensing base station are obtained based on the training samples at each execution step. The policy generation unit is trained using the following steps: For the current execution step, determine the target value of the value function subnetwork corresponding to the current execution step. The target value is used to indicate the benefit from the current execution step to the termination execution step during the approach and landing process of the test aircraft in this round of training. The temporal difference target value is determined based on the training samples at the current execution step, the test sensing base station execution strategy and the reward value, and the training samples at the next execution step; The error loss is determined based on the target value of the value function subnetwork corresponding to the current execution step and the temporal difference target value; The policy generation unit is trained by minimizing the error loss.

[0011] In one optional implementation, determining the temporal difference target value based on the training samples at the current execution step, the test sensing base station execution strategy and the reward value, and the training samples at the next execution step includes: Based on the training samples, the test sensing base station execution strategy, and the reward value at the current execution step, and the training samples at the next execution step, the predicted test sensing base station execution strategy for the next execution step is predicted. Based on the predicted test sensing base station execution strategy under the next execution step, predict the prediction value of the value function sub-network corresponding to the next execution step. The prediction value is used to indicate the benefit between the next execution step and the termination execution step during the approach and landing process of the test aircraft under this round of training. The temporal difference target value is determined based on the reward value at the current execution step and the prediction value of the value function subnetwork corresponding to the next execution step.

[0012] According to a second aspect of this application, an aircraft approach and landing assistance device is provided, the device comprising: The data acquisition module is used to acquire radar point cloud data of aircraft collected by the sensing base station during the approach and landing of multiple aircraft; the sensing base station is set at the top of the multi-layer rotating vertical take-off and landing field, which includes multiple externally suspended take-off and landing platforms. The strategy prediction module is used to input the radar point cloud data of the aircraft into a pre-trained strategy prediction model to obtain the sensor base station execution strategy output by the strategy prediction model. The sensor base station execution strategy is used to instruct the sensor base station to perform actions in the process of providing communication services and landing alignment services to the aircraft. The strategy execution module is used to execute the strategy of the sensing base station, control the sensing base station to communicate with the multiple aircraft, and control the sensing base station to adjust the multiple externally suspended take-off and landing platforms.

[0013] According to a third aspect of this application, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the aircraft approach and landing assistance method described in the first aspect above.

[0014] According to a fourth aspect of this application, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the aircraft approach and landing assistance method described in the first aspect above.

[0015] The aircraft approach and landing assistance method, apparatus, equipment, and medium provided in this application, during the approach and landing of multiple aircraft, collects aircraft radar point cloud data through a sensing base station. The sensing base station is located at the top of a multi-layer rotating vertical take-off and landing field, which includes multiple externally suspended take-off and landing platforms. An execution strategy for the sensing base station is generated based on the aircraft radar point cloud data to control communication between the sensing base station and multiple aircraft, and to control the sensing base station to adjust the multiple externally suspended take-off and landing platforms. This multi-layer rotating vertical take-off and landing field helps reduce the land area occupied and increases the aircraft's carrying capacity. Compared to a single-layer or planar layout, it significantly improves land use efficiency. Furthermore, the sensing base station can adjust the multiple externally suspended take-off and landing platforms, allowing them to simultaneously align with the aircraft's attitude during approach and landing, enabling the aircraft to quickly enter the descent phase, reducing the time the aircraft spends hovering in the airspace, lowering propulsion energy consumption, operating costs, and safety hazards caused by low-altitude congestion, and improving the safety and efficiency of aircraft approach and landing.

[0016] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the technical solutions of this disclosure.

[0017] To make the above-mentioned objects, features and advantages of this disclosure more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating an exemplary embodiment of an aircraft approach and landing assistance method. Figure 2 This is a schematic diagram of a sensing base station and a multi-layer rotating vertical take-off and landing field shown in an exemplary embodiment of this application; Figure 3 This is a schematic diagram illustrating the structure of a strategy prediction model according to an exemplary embodiment of this application; Figure 4 This is a schematic diagram of an aircraft approach and landing assistance device shown in an exemplary embodiment of this application; Figure 5 This is a schematic diagram of the structure of a computer device shown in an exemplary embodiment of this application. Detailed Implementation

[0019] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0020] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0021] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0022] Research has revealed that aircraft approach, queuing, and landing in urban low-altitude airspace rely on ground-based vertical takeoff and landing (Vertiports). To meet safety clearance and landing disk diameter requirements, Vertiport deployment typically requires significant horizontal space, resulting in large footprints, low land utilization, and difficulty in landing in core urban areas. Furthermore, traditional Vertiports employ a single-plane layout with a fixed structure, limiting the number of aircraft they can accommodate simultaneously and hindering capacity expansion, making them unsuitable for the high-density operations required by Urban Aircraft Management (UAM). Due to the limited number of takeoff and landing platforms, when multiple aircraft approach simultaneously, they often need to hover at low altitudes and spend considerable time adjusting their attitude to ensure alignment with the platforms, lacking queuing buffer capacity, leading to increased energy consumption and safety risks. To improve aircraft approach and landing efficiency, most methods employ altitude-based queuing, neglecting the scalability of the Vertiport itself.

[0023] Based on the above research, this application provides an aircraft approach and landing assistance method, device, equipment, and medium. It adopts a multi-layer rotating vertical take-off and landing field, which helps to reduce the footprint, increase the aircraft's reception capacity, and can adjust multiple externally suspended take-off and landing platforms through a sensing base station. This allows multiple externally suspended take-off and landing platforms to simultaneously coordinate with the aircraft's attitude during approach and landing, align with and receive the aircraft, enabling the aircraft to quickly enter the descent phase, reduce the time the aircraft spends hovering in the airspace, and reduce propulsion energy consumption, operating costs, and safety hazards caused by low-altitude congestion.

[0024] To facilitate understanding of this embodiment, a detailed description of the aircraft approach and landing assistance method disclosed in this application embodiment will be provided first. The executing entity of the aircraft approach and landing assistance method provided in this application embodiment is generally an electronic device with a certain computing power. This electronic device can be a server, which can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud storage, big data, and artificial intelligence platforms. In some possible implementations, the aircraft approach and landing assistance method can be implemented by a processor calling computer-readable instructions stored in memory.

[0025] The following description, in conjunction with the accompanying drawings, illustrates an aircraft approach and landing assistance method provided in an embodiment of this application.

[0026] See Figure 1 The diagram shown is a flowchart illustrating an exemplary embodiment of this application for an aircraft approach and landing assistance method. Figure 1 As shown in the figure, the aircraft approach and landing assistance method provided in this embodiment includes steps S101 to S103, wherein: S101: During the approach and landing of multiple aircraft, acquire radar point cloud data of the aircraft collected by the sensing base station; the sensing base station is set at the top of the multi-layer rotating vertical take-off and landing field, which includes multiple externally suspended take-off and landing platforms.

[0027] In this step, during the approach and landing of multiple aircraft, a sensor base station can be used to assist the aircraft in approach and landing. Specifically, the sensor base station can be used to provide communication services and landing alignment services for the aircraft.

[0028] For example, in low-altitude areas (e.g., below 3 kilometers), multiple aircraft descend step by step from the outskirts of a city or high-altitude areas along a predetermined approach corridor to the corresponding take-off and landing platforms.

[0029] Here, the aircraft referred to is an aircraft with approach capabilities, such as an electric vertical take-off and landing (eVTOL) aircraft or a general aviation aircraft. This disclosure uses an eVTOL as an example for illustration. Relying on electric propulsion, multi-rotor, or tiltrotor structures, eVTOLs possess significant advantages such as zero emissions, low noise, low vibration, vertical take-off and landing, high maneuverability, and high intelligence. They can perform diverse tasks in urban environments, including short-distance, on-demand travel, emergency rescue, medical transport, and logistics delivery. Compared to traditional helicopters, eVTOLs have advantages such as low operating costs, simple maintenance, and high safety, and are widely considered one of the representative and feasible core vehicles for urban low-altitude transportation.

[0030] The sensing base station includes millimeter-wave (mmWave) antenna arrays, radar units, etc., and can generate radar point cloud (RPC) data in real time to sense the position and motion status of surrounding aircraft, providing trajectory monitoring and environmental awareness for approaching and landing aircraft. Simultaneously, the aircraft can maintain data interaction with the sensing base station during approach to complete status monitoring, mission data uploading, and control command issuance. This embodiment employs Integrated Sensing and Communication (ISAC) technology, which, by sharing radar and communication spectrum resources, achieves joint processing of sensing information and communication signals, significantly improving spectrum utilization and system coordination performance.

[0031] The sensing base station is located at the top of the multi-layer rotating vertical takeoff and landing (VTOL) field. The sensing beams can be finely adjusted via the tower's rotating structure to ensure they always cover the main approach path of the aircraft, thereby improving signal stability and target identification accuracy. In high-density access scenarios, the sensing base station can acquire information such as landing priority, remaining battery power, altitude, and position of each aircraft, thus determining the adjustment amount and access sequence for different externally mounted platforms. This allows the overall structure formed by the sensing base station and the multi-layered VTOL field to function as a takeoff and landing node, a communication node, and a sensing node, becoming a core hub in the urban low-altitude network and contributing to improved scalability and reliability of aircraft approach and landing.

[0032] See Figure 2 This is a schematic diagram illustrating a sensing base station and a multi-layer rotating vertical takeoff and landing field, as shown in an exemplary embodiment of this application. Figure 2As shown, a sensing base station 20 is installed on top of the multi-level rotating vertical takeoff and landing field 10. This embodiment of the present disclosure uses a sensing base station installed on top of the multi-level rotating vertical takeoff and landing field to provide approach guidance for multiple approaching and landing aircraft based on integrated sensing and communication technology. On the one hand, it acquires aircraft radar point cloud data through millimeter-wave radar; on the other hand, it establishes a short-range air-to-ground link to issue approach commands. Combined with the geometric position and attitude angle of each externally mounted takeoff and landing platform, it completes closed-loop control of platform selection, attitude alignment, and approach guidance, thereby reducing the costs of aircraft holding in the air and terminal maneuver adjustments.

[0033] The multi-layer rotating vertical take-off and landing (VTOL) field 10 is a multi-layer tower structure, comprising a central load-bearing tower column 11, multiple layered support arms 12, and multiple externally suspended take-off and landing platforms 13. The multi-layer rotating vertical take-off and landing field 10 can be expanded into multiple layers; the specific number of layers can be determined according to actual needs and is not limited here. Under each layer, multiple layered support arms 12 extend outwards; the specific number of layered support arms can be determined according to actual needs and is not limited here. The layered support arms 12 can have a tree-like or umbrella-like structure, with each layered support arm 12 connecting to an externally suspended take-off and landing platform 13, thereby enabling the simultaneous arrangement of several externally suspended take-off and landing platforms 13 in multiple directions, forming a high-density vertically stacked take-off and landing combination. The multi-layer rotating vertical take-off and landing field 10 can integrate power supply lines, rotating drive shafts, sensor lines, and maintenance channels, allowing adjustments to each externally suspended take-off and landing platform 13 to be made independently without affecting the overall structure of the multi-layer rotating vertical take-off and landing field 10. Through this high-level integrated structure, the embodiments of this disclosure can realize the spatial stacking of multiple take-off and landing platforms with a very small footprint, greatly improving the air traffic handling capacity per unit land area and enhancing the feasibility of deploying take-off and landing fields in urban core areas.

[0034] Each layer of the overhanging take-off and landing platform 13 adopts an overhanging multi-branch structure, including multiple layered support arms 12 and multiple overhanging take-off and landing platforms 13. The specific dimensions of each overhanging take-off and landing platform 13 can be designed according to the aircraft's wingspan, rotor diameter, and safety clearance requirements, and are equipped with anti-skid take-off and landing surfaces and night / inclement weather guidance lighting. The branch structure can be deployed in multiple directions, such as "semi-tree," "Y-shaped," or "multi-radial," allowing multiple independent take-off and landing positions to be accommodated in the same layer, forming a multi-aircraft simultaneous access capability. The deployment method of the overhanging take-off and landing platform 13 supports modular installation, and different numbers of layered support arms 12 can be flexibly configured according to the urban building environment or the needs of three-dimensional transportation, realizing expandable and relocatable deployment. Compared with the traditional Vertiport, the embodiments disclosed in this disclosure can reduce the dependence on large-area planar sites. Through vertical capacity expansion, the airspace processing capacity is no longer limited to a single plane, which helps to improve the problem of insufficient deployability caused by land scarcity in urban centers.

[0035] Each layer also includes a horizontal rotation mechanism 14, which enables the left and right rotation of each externally suspended take-off and landing platform 13. Specifically, it enables each externally suspended take-off and landing platform 13 to rotate 360° horizontally around the central load-bearing tower 11. The horizontal rotation mechanism 14 may include components such as a circular rotating track, a geared motor, a slewing bearing, a torque limiting system, and an angle encoder. Driven by the motor, the platform orientation can be adjusted arbitrarily within a 360° omnidirectional angle range. Under the control of the sensing base station 20, the externally suspended take-off and landing platform 13 can be automatically rotated to the optimal docking direction according to the direction of the aircraft's approach, so that the aircraft's approach track is consistent with the centerline of the externally suspended take-off and landing platform 13, reducing the need for the aircraft to perform complex aerial maneuvers or large-angle turns to match the fixed platform orientation of the externally suspended take-off and landing platform 13. Based on the radar point cloud data of the aircraft collected by the sensing base station 20, the rotation angle of each externally suspended take-off and landing platform 13 can be automatically controlled, so that aircraft approaching from different directions can land directly on the most suitable externally suspended take-off and landing platform 13. The real-time orientation alignment capability of the horizontal rotation mechanism 14 enables a single externally suspended take-off and landing platform 13 to serve aircraft approaching from different directions multiple times after rotation, effectively improving the take-off and landing field capacity and scheduling efficiency.

[0036] In addition to the horizontal rotation mechanism 14, each layer also includes a pitch movement mechanism 15, enabling each externally suspended take-off and landing platform 13 to move vertically in the vertical plane. The pitch movement mechanism 15 may include a pitch pivot, hydraulic or servo-electric actuators, angle sensors, and a stability control system. The pitch angle of the externally suspended take-off and landing platform 13 can be adjusted via the control of the sensing base station 20. Based on the aircraft radar point cloud data collected by the sensing base station 20, the pitch angle of the externally suspended take-off and landing platform 13 can be actively adjusted so that the platform faces the direction of the aircraft's descent, thereby shortening the alignment distance, reducing the attitude adjustments required at the end of the approach and landing process, and improving landing safety. During simultaneous landings of multiple aircraft, the vertical movement of each externally suspended take-off and landing platform 13 allows for staggered access, improving space utilization efficiency and enhancing landing performance and safety.

[0037] The multi-layered rotating vertical takeoff and landing (VTOL) field employs a dual-degree-of-freedom structural design, enabling 360° horizontal omnidirectional rotation and controllable pitch vertical movement. This allows each platform to actively align itself with the aircraft's position and angle. Compared to traditional methods that require aircraft to adjust their attitude, hover in queue, or maneuver for alignment, this mechanism, by having platforms actively approach the aircraft, significantly reduces the time aircraft spend in the air waiting and attitude adjustment, improves access efficiency, and significantly reduces energy consumption and safety risks. It is particularly suitable for urban core areas with restricted airspace, narrow corridors, and complex wind conditions.

[0038] Here, the multi-layered rotating vertical takeoff and landing (VTOL) field features multiple layered arms and externally suspended takeoff and landing platforms at each level, enabling multiple aircraft to simultaneously approach at each altitude. Combined with the multi-layered vertical stacking, this creates a high-density access space, significantly alleviating the severe queuing problems caused by the traditional single-entry, single-exit approach of Vertiport. By employing a three-dimensional, multi-layered tower structure, multiple takeoff and landing platforms are compactly stacked vertically, combined with a tree-like branching layout, allowing the takeoff and landing field, which originally required a large area of ​​ground space, to achieve a multiple-fold capacity expansion within a minimal footprint. The tree-like structure not only enhances capacity but also allows for dynamic scheduling based on the number of aircraft, preventing mutual interference between externally suspended takeoff and landing platforms, significantly increasing the overall throughput for approach, docking, and departure, and reducing low-altitude hovering congestion.

[0039] The sensing base station and multi-layer rotating vertical take-off and landing field provided in this disclosure embodiment, by adopting a multi-layer rotating vertical take-off and landing field with a multi-layer branch structure, can achieve dual degrees of freedom of left and right rotation and up and down movement. This allows each external rotating take-off and landing platform to automatically align with the approach direction and traffic load of the aircraft. It has a small footprint, high scalability, and can serve multiple aircraft simultaneously, thereby improving land utilization, increasing the capacity of the take-off and landing field, enabling simultaneous access from multiple directions, improving aircraft approach efficiency, reducing the time for aircraft to hover or circle in the air and wait in line, thereby increasing the UAM access capacity in urban areas and ensuring the safe operation of high-density aircraft traffic.

[0040] Optionally, the multi-layer rotating vertical take-off and landing field in this embodiment adopts a modular structural design. Each layer, each platform, horizontal rotation mechanism, and pitch movement mechanism can be manufactured in a standardized manner and assembled on-site, facilitating maintenance, expansion, and upgrades. The modularity enables customized configurations for different cities, plots, and capacity requirements, and can be gradually expanded vertically or horizontally as traffic demand grows, improving project feasibility.

[0041] S102: Input the radar point cloud data of the aircraft into the pre-trained strategy prediction model to obtain the sensor base station execution strategy output by the strategy prediction model. The sensor base station execution strategy is used to instruct the sensor base station to perform actions in the process of providing communication services and landing alignment services to the aircraft.

[0042] In this step, the aircraft radar point cloud data can be attributed to a pre-trained strategy prediction model. The strategy prediction model can then generate a corresponding sensing base station execution strategy based on the aircraft radar point cloud data.

[0043] Please see Figure 3 This is a schematic diagram illustrating the structure of a policy prediction model, as shown in an exemplary embodiment of this application. Figure 3As shown, the policy prediction model includes a radar point cloud feature extraction unit and a policy generation unit.

[0044] In some possible implementations, the step of inputting the aircraft radar point cloud data into a pre-trained policy prediction model to obtain the sensing base station execution policy output by the policy prediction model includes: The radar point cloud feature extraction unit extracts features from the aircraft radar point cloud data to obtain a radar point cloud feature vector. The aircraft radar point cloud data includes the aircraft position, the aircraft azimuth angle, the aircraft pitch angle, the relative distance between the sensing base station and the aircraft, the relative speed between the sensing base station and the aircraft, and the radar signal-to-noise ratio of the aircraft. The strategy generation unit performs strategy reasoning based on the radar point cloud feature vector to obtain the execution strategy of the sensing base station; the execution strategy of the sensing base station includes the communication power allocation ratio between the sensing base station and each aircraft, the horizontal rotation control amount of each externally suspended take-off and landing platform, and the pitch control amount of each externally suspended take-off and landing platform.

[0045] Here, the sensing base station uses millimeter-wave radar to continuously collect information about surrounding aircraft and generate radar point cloud data of the aircraft.

[0046] The radar point cloud data of the aircraft includes the aircraft's position, azimuth angle, pitch angle, relative distance between the sensing base station and the aircraft, relative velocity between the sensing base station and the aircraft, and radar cross-section between the sensing base station and the aircraft.

[0047] Specifically, the aircraft position can include the aircraft's position along the horizontal, vertical, and lateral axes. For example, the aircraft position can be represented by the following formula (1): (1) in, This represents the three-dimensional position of the k-th eVTOL at time t. This represents the x-axis coordinate of the k-th eVTOL at time t. This represents the ordinate of the k-th eVTOL at time t. Let represent the vertical coordinate of the k-th eVTOL at time t.

[0048] The azimuth angle of the aircraft can be determined based on the position of the eVTOL in the horizontal axis direction, the position of the eVTOL in the vertical axis direction, the position of the sensing base station in the horizontal axis direction, and the position of the sensing base station in the vertical axis direction.

[0049] For example, the location of the sensing base station can be represented by the following formula (2): (2) in, Indicates the three-dimensional location of the sensing base station. The horizontal axis coordinates of the sensor base station are represented. The vertical axis coordinates of the sensing base station are represented. This represents the vertical axis coordinates of the sensor base station.

[0050] The azimuth angle of the aircraft can be determined by the following formula (3): (3) in, This represents the azimuth angle of the k-th eVTOL relative to the sensing base station at time t. This represents the ordinate of the k-th eVTOL at time t. The vertical axis coordinates of the sensing base station are represented. This represents the x-axis coordinate of the k-th eVTOL at time t. The horizontal axis coordinates of the sensor base station are represented. This represents the arctangent function.

[0051] The pitch angle of the aircraft can be determined based on the positions of the eVTOL in the horizontal axis, the vertical axis, and the vertical direction, as well as the positions of the sensing base station in the horizontal axis, the vertical axis, and the vertical direction. For example, the pitch angle of the aircraft can be determined using the following formula (4): (4) in, This represents the pitch angle of the k-th eVTOL relative to the sensing base station at time t. This represents the vertical coordinate of the k-th eVTOL at time t. Represents the vertical axis coordinates of the sensor base station. This represents the x-axis coordinate of the k-th eVTOL at time t. The horizontal axis coordinates of the sensor base station are represented. This represents the ordinate of the k-th eVTOL at time t. The vertical axis coordinate represents the coordinates of the sensing base station.

[0052] The relative distance between the sensing base station and the aircraft can be determined based on the location of the aircraft and the location of the sensing base station. For example, the relative distance can be determined using the following formula (5): (5) in, This represents the relative distance between the k-th eVTOL and the sensing base station at time t. This represents the three-dimensional position of the k-th eVTOL at time t. Indicates the three-dimensional location of the sensing base station. This represents the L2 norm.

[0053] The relative speed between the sensing base station and the aircraft can be determined by dividing the relative distance by the time.

[0054] The radar signal-to-noise ratio of the aircraft can be determined based on the echo reception power of the sensor base station to the aircraft, the radar received noise spectral density, and the radar bandwidth. The echo reception power of the sensor base station to the aircraft can be determined based on the radar transmission power of the sensor base station, the radar transmission antenna gain of the sensor base station, the radar receiving antenna gain of the aircraft, the radar carrier wavelength, the radar cross section (RCS) of the aircraft, and the relative distance.

[0055] For example, the echo reception power of the sensing base station to the aircraft can be determined by the following formula (6): (6) in, This represents the echo reception power of the sensing base station for the k-th eVTOL at time t. Indicates the radar transmit power of the sensing base station. This indicates the gain of the radar transmitting antenna of the sensing base station. Indicates the gain of the receiving antenna of the k-th eVTOL radar. Indicates the radar carrier wavelength. This represents the radar cross-section of an aircraft. This represents the relative distance between the k-th eVTOL and the sensing base station at time t.

[0056] The radar signal-to-noise ratio of the aircraft can be determined by the following formula (7): (7) in, This represents the radar signal-to-noise ratio of the k-th eVTOL at time t. This represents the echo reception power of the sensing base station for the k-th eVTOL at time t. This represents the radar received noise spectral density. Indicates radar bandwidth.

[0057] In this way, the radar point cloud feature extraction unit condenses the aircraft's position, angle, distance, speed, and signal-to-noise ratio into a unified feature vector. Based on this, the strategy generation unit determines the communication power allocation ratio, platform horizontal rotation, and pitch control quantities. The sensing base station can enable multiple externally suspended take-off and landing platforms to match the approach attitude of multiple aircraft in real time without repeated attempts, achieving precise alignment, stable communication, and rapid reception under high-density take-off and landing, significantly reducing low-altitude waiting time, reducing energy consumption and interference, and improving overall operational efficiency and safety margin.

[0058] In some possible implementations, the policy prediction model is obtained by training a reinforcement learning network: A state space representing the environmental information of the aircraft's approach and landing is constructed. Based on the state space, training samples are generated, including radar point cloud data of the test aircraft collected by the test sensing base station. The training samples are input into the reinforcement learning network to obtain the test sensory base station execution strategy output by the reinforcement learning network. When executing the test sensing base station execution strategy, determine the platform selection cost corresponding to each test aircraft choosing to land on an externally suspended take-off and landing platform, and determine the reward value corresponding to the test sensing base station execution strategy; Minimizing the platform selection cost is determined as the optimization objective. The reinforcement learning network is iteratively trained until the network meets the cutoff condition based on the reward value. The reinforcement learning network that meets the cutoff condition is then determined as the trained policy prediction model.

[0059] In the above steps, a state space representing the environmental information of the aircraft's approach and landing can be constructed. This state space includes the aircraft's direction of arrival, communication link, and platform attitude. Specifically, the state space can be represented by the following formula (8): (8) in, This represents the system state at time t. This represents the three-dimensional position of the k-th test eVTOL at time t, where K represents the total number of test eVTOLs. This represents the azimuth angle of the k-th test eVTOL relative to the test sensing base station at time t. This represents the pitch angle of the k-th test eVTOL relative to the test sensing base station at time t. This represents the communication channel capacity between the test sensing base station and the k-th test eVTOL at time t. This represents the horizontal rotation angle of the m-th externally suspended take-off and landing platform at time t. Let represent the pitch angle of the m-th externally suspended take-off and landing platform at time t, and M represent the total number of externally suspended take-off and landing platforms.

[0060] Based on the state space, training samples are generated, including radar point cloud data of the test aircraft collected by the test sensing base station. The training samples are input into the reinforcement learning network to obtain the test sensing base station execution strategy output by the neural network. The test sensing base station execution strategy includes the communication power allocation ratio between the test sensing base station and each test aircraft, the horizontal rotation control amount of each externally mounted take-off and landing platform, and the pitch control amount of each externally mounted take-off and landing platform.

[0061] For example, the action space corresponding to the reinforcement learning network can be represented by the following formula (9): (9) in, Indicates the action at time t. This represents the communication power allocation ratio between the test sensing base station and the k-th test eVTOL at time t, where K represents the total number of test eVTOLs. This represents the horizontal rotation control value of the m-th externally suspended take-off and landing platform at time t. Let M represent the pitch control value of the m-th externally suspended take-off and landing platform at time t, and M represent the total number of externally suspended take-off and landing platforms.

[0062] The communication power allocation ratio needs to meet the normalization constraint. Specifically, the communication power allocation ratio between the test sensing base station and each test aircraft is greater than 0 and less than 1, and the sum of the communication power allocation ratios of the multiple test aircraft is 1. This can be expressed by the following formula (10): (10) in, This represents the communication power between the test sensing base station and the k-th test eVTOL at time t. This represents the communication power allocation ratio between the test sensing base station and the k-th test eVTOL at time t. This indicates the total communication power of the tested sensing base station. This represents the sum of the communication power allocation ratios between the test sensing base station and multiple test eVTOLs at time t, where K represents the total number of test eVTOLs.

[0063] When executing the test sensing base station execution strategy, the platform selection cost corresponding to the landing on the externally suspended take-off and landing platform for each test aircraft is determined, as well as the reward value corresponding to the test sensing base station execution strategy. Minimizing the platform selection cost is set as the optimization objective, and the reinforcement learning network is iteratively trained until the reinforcement learning network meets the cutoff condition based on the reward value. The reinforcement learning network that meets the cutoff condition is then determined as the trained policy prediction model.

[0064] In this way, the reinforcement learning network optimizes by minimizing the cost of platform selection. During iterative training, it autonomously discovers the optimal mapping relationship between the aircraft state and platform resources. The trained policy prediction model can directly output the communication power allocation ratio and platform attitude control commands, enabling the sensing base station to allocate the most suitable external landing platform to each aircraft in real time in high-density approach scenarios. This significantly reduces waiting time, energy consumption, and conflict risks, achieving autonomous landing guidance with simultaneous improvement in safety margin and operational efficiency. Furthermore, the external landing platform can complete orientation and docking preparation as soon as the aircraft enters the approach phase. Compared to traditional static landing platforms, it does not rely on complex aircraft maneuvers to adapt to the platform orientation. Instead, it weakens crosswinds, adjusts alignment angles, and avoids inter-platform motion interference through the platform's own movement, achieving a higher level of operational safety and access efficiency.

[0065] In some possible implementations, the platform selection cost corresponding to landing on an externally suspended takeoff and landing platform for any of the test aircraft is determined by the following steps: Based on the position of the test aircraft and the position of the externally suspended take-off and landing platform selected by the test aircraft for landing, the distance error between the test aircraft and the corresponding externally suspended take-off and landing platform is determined; Based on the azimuth angle of the test aircraft and the corresponding horizontal rotation angle of the externally suspended take-off and landing platform, the horizontal alignment error between the test aircraft and the corresponding externally suspended take-off and landing platform is determined. Based on the pitch angle of the test aircraft and the corresponding pitch angle of the externally suspended take-off and landing platform, determine the pitch direction alignment error between the test aircraft and the corresponding externally suspended take-off and landing platform; Based on the communication bandwidth between the test aircraft and the corresponding external take-off and landing platform and the signal-to-noise ratio of the communication link of the test aircraft, the communication channel capacity between the test aircraft and the corresponding external take-off and landing platform is determined. Based on the distance error, horizontal alignment error, pitch alignment error, and communication channel capacity between the test aircraft and the corresponding externally suspended take-off and landing platform, the platform selection cost corresponding to the test aircraft choosing the externally suspended take-off and landing platform for landing is determined.

[0066] In the above steps, let the center position of the m-th externally suspended take-off and landing platform be... Its horizontal rotation angle is Its pitch angle is Specifically, it can be expressed by the following formula (11): (11) in, This represents the three-dimensional position of the center of the m-th externally suspended take-off and landing platform. This represents the horizontal rotation angle of the m-th externally suspended take-off and landing platform, which can rotate freely left and right within 360°. This represents the pitch angle of the m-th externally suspended take-off and landing platform. Indicates the permissible pitch range.

[0067] Based on the difference between the position of the test aircraft and the position of the externally suspended take-off and landing platform selected by the test aircraft for landing, the distance error between the test aircraft and the corresponding externally suspended take-off and landing platform is determined. Specifically, the distance error can be determined by the following formula (12): (12) in, This represents the distance error between the k-th test eVTOL and the m-th externally suspended take-off and landing platform at time t. This represents the three-dimensional position of the k-th test eVTOL at time t. This represents the three-dimensional position of the center of the m-th externally suspended take-off and landing platform.

[0068] Based on the difference between the azimuth angle of the test aircraft and the horizontal rotation angle of the corresponding externally suspended take-off and landing platform, the horizontal alignment error between the test aircraft and the corresponding externally suspended take-off and landing platform is determined. Specifically, the horizontal alignment error can be determined by the following formula (13): (13) in, This represents the horizontal alignment error between the k-th test eVTOL and the m-th externally suspended take-off and landing platform at time t. This represents the azimuth angle of the k-th test eVTOL relative to the test sensing base station at time t. This represents the horizontal rotation angle of the m-th externally suspended take-off and landing platform at time t.

[0069] Based on the difference between the pitch angle of the test aircraft and the pitch angle of the corresponding externally suspended take-off and landing platform, the pitch alignment error between the test aircraft and the corresponding externally suspended take-off and landing platform is determined. Specifically, the pitch alignment error can be determined by the following formula (14): (14) in, This represents the pitch alignment error between the k-th test eVTOL and the m-th externally suspended take-off and landing platform at time t. This represents the pitch angle of the k-th test eVTOL relative to the test sensing base station at time t. This represents the horizontal rotation angle of the m-th externally suspended take-off and landing platform at time t.

[0070] Based on the communication bandwidth between the test aircraft and the corresponding externally suspended takeoff and landing platform, and the signal-to-noise ratio (SNR) of the communication link between the test aircraft and the corresponding externally suspended takeoff and landing platform, the communication channel capacity between the test aircraft and the corresponding externally suspended takeoff and landing platform is determined. Here, the SNR of the communication link between the test aircraft and the test sensor base station can be determined based on the communication power of the test sensor base station to the test aircraft, the gain of the test sensor base station's communication transmitting antenna, the gain of the test aircraft's communication receiving antenna, the communication noise spectral density, the communication bandwidth, and the channel gain. The channel gain can be determined based on the communication link loss of the test aircraft. The communication link loss of the test aircraft can be determined based on the reference distance loss constant, the relative distance between the test sensor base station and the test aircraft, and the shadowing fading of the test aircraft.

[0071] Specifically, the communication link loss of the test aircraft can be determined by the following formula (15): (15) in, This represents the link loss of the k-th test eVTOL at time t. Reference distance loss constant, This represents the path loss index. This represents the relative distance between the k-th test eVTOL and the test sensing base station at time t. Let t represent the shadow fading of the k-th test eVTOL at time t. The shadow fading can be specifically modeled as a Gaussian random variable.

[0072] The link loss is converted to a linear-scale channel gain. Specifically, the channel gain can be determined by the following formula (16): (16) in, This represents the channel gain of the k-th test eVTOL at time t. Let represent the link loss of the k-th test eVTOL at time t.

[0073] The signal-to-noise ratio of the communication link of the test aircraft can be determined by the following formula (17): (17) in, This represents the signal-to-noise ratio of the communication link of the k-th test eVTOL at time t. This represents the communication power between the test sensing base station and the k-th test eVTOL at time t. This indicates the gain of the transmitting antenna for the inductive base station communication test. This represents the gain of the eVTOL communication receiving antenna of the k-th test unit. This represents the channel gain of the k-th test eVTOL at time t. Represents the communication noise spectral density. Indicates communication bandwidth.

[0074] The communication channel capacity can be determined by the following formula (18): (18) in, This represents the communication channel capacity of the k-th test eVTOL at time t. Indicates communication bandwidth. Let represent the signal-to-noise ratio of the communication link of the k-th test eVTOL at time t.

[0075] Based on the distance error, horizontal alignment error, pitch alignment error, and communication channel capacity between the test aircraft and the corresponding external take-off and landing platform, the platform selection cost corresponding to the test aircraft choosing the external take-off and landing platform for landing is determined. Specifically, the platform selection cost can be determined by the following formula (19): (19) in, This represents the platform selection cost between the k-th test eVTOL and the m-th externally suspended take-off and landing platform at time t. Indicates the weighting coefficient. This represents the distance error between the k-th test eVTOL and the m-th externally suspended take-off and landing platform at time t. This represents the horizontal alignment error between the k-th test eVTOL and the m-th externally suspended take-off and landing platform at time t. This represents the pitch alignment error between the k-th test eVTOL and the m-th externally suspended take-off and landing platform at time t. Let represent the communication channel capacity of the k-th test eVTOL at time t. Here, The larger the value, the more likely it is to be selected, hence the use of a negative sign.

[0076] In this way, the platform selection cost comprehensively considers distance error, horizontal alignment error, pitch alignment error, and communication channel capacity, and comprehensively evaluates the deviations between the test aircraft and the corresponding external take-off and landing platform in terms of spatial position, attitude matching, and information transmission capabilities. This enables the trained strategy prediction model to prioritize guiding the aircraft to the external take-off and landing platform that is closest, has the best attitude matching, and has the most stable communication. This allows for precise alignment, stable communication, and efficient landing, significantly improving operational efficiency, safety margin, and resource utilization in high-density approach scenarios.

[0077] Alternatively, the optimal platform selection decision can be determined by minimizing the platform selection cost, which can be expressed by the following formula (20): (20) in, This indicates that at time t, the externally suspended takeoff and landing platform selected for the k-th test eVTOL is... This represents the set of all available externally suspended take-off and landing platforms, and argmin represents the value of the independent variable for which the function reaches its minimum value. Let represent the platform selection cost between the k-th test eVTOL and the m-th externally suspended takeoff and landing platform at time t. It can be understood that, through formula (20), the cost at time t can be determined... The number of the smallest externally suspended take-off and landing platform.

[0078] In some possible implementations, the reinforcement learning network is determined to satisfy the cutoff condition through the following steps: Based on the communication channel capacity, distance error, horizontal alignment error, pitch alignment error, horizontal rotation control amount and pitch control amount of each externally suspended take-off and landing platform between the test aircraft and the corresponding externally suspended take-off and landing platform, the reward value corresponding to the execution strategy of the test sensing base station is determined. When the reward value tends to stabilize, the reinforcement learning network is determined to meet the cutoff condition.

[0079] Here, in the joint control of the sensor-guided base station and the multi-layer rotating vertical take-off and landing field, the system state includes the three-dimensional positions, approach angles, communication link states of multiple eVTOLs, and attitude information of multiple externally suspended take-off and landing platforms. Both the state space and action space exhibit continuous, high-dimensional, and strongly coupled characteristics, making it difficult to solve in real-time in dynamic urban low-altitude environments using traditional rule-based or analytical optimization methods. Therefore, this disclosure employs a deep reinforcement learning method based on an Actor-Critic structure to jointly learn and update the sensor-guided, power allocation, and platform rotation control strategies online.

[0080] The policy generation unit comprises a policy sub-network (Actor) and a value function sub-network (Critic), with the Actor and Critic networks performing different but complementary functions. The Actor network outputs the execution policy of the test sensing base station, with training samples as its input. The Actor network directly determines the system's control behavior at the current moment, aiming to maximize the cumulative reward value over long-term operation, thereby improving overall approach efficiency, communication capacity, and platform alignment performance. The Critic network evaluates the value of the test sensing base station's execution policy, generating the reward value corresponding to the policy—that is, the expected long-term return obtainable after executing the action (i.e., the test sensing base station execution policy) in the current state. Its input is a state-action pair. The Critic network does not directly participate in control decisions but provides directional guidance for the Actor network's parameter updates by estimating the reward value of the test sensing base station's execution policy, enabling the Actor to gradually correct its policy and converge towards better control behavior.

[0081] Specifically, the reward value corresponding to the execution strategy of the test sensing base station can be determined based on the communication channel capacity, distance error, horizontal alignment error, pitch alignment error, and the horizontal rotation control and pitch control of each externally suspended take-off and landing platform between the test aircraft and the corresponding externally suspended take-off and landing platform. For example, the reward value can be determined by the following formula (21): (twenty one) in, This represents the reward value at time t. Indicates the weighting coefficient. This represents the communication channel capacity of the k-th test eVTOL at time t. This refers to the externally suspended takeoff and landing platform used for the test eVTOL of aircraft k. This indicates that at time t, the k-th test eVTOL and the corresponding selected eVTOL... Distance error between externally suspended take-off and landing platforms This indicates that at time t, the k-th test eVTOL and the corresponding selected eVTOL... Horizontal alignment error between the externally suspended take-off and landing platforms This indicates that at time t, the k-th test eVTOL and the corresponding selected eVTOL... Pitch alignment error between the externally suspended take-off and landing platforms This represents the horizontal rotation control value of the m-th externally suspended take-off and landing platform at time t. Let M represent the pitch control value of the m-th externally suspended take-off and landing platform at time t, and M represent the total number of externally suspended take-off and landing platforms.

[0082] In this way, the reward value comprehensively considers the communication channel capacity, three-dimensional errors (distance error, horizontal alignment error, and pitch alignment error) and platform control variables (horizontal rotation control variable and pitch control variable). When the reward value tends to stabilize during training, the network is considered to have converged. The resulting policy prediction model can stably output a sensing base station execution policy that takes into account information quality, spatial accuracy, and energy consumption. This enables the sensing base station to maintain accurate alignment, reliable communication, and low-power operation during high-density approaches, significantly improving the overall landing efficiency, safety margin, and resource utilization.

[0083] In some possible implementations, the policy generation unit includes a policy subnetwork and a value function subnetwork. The policy subnetwork is used to output the execution policy of the test sensing base station, and the value function subnetwork is used to evaluate the value of the execution policy of the test sensing base station and generate a reward value corresponding to the execution policy of the test sensing base station. Each training round includes multiple consecutive execution steps, and the corresponding execution policy and reward value of the test sensing base station are obtained based on the training samples at each execution step. The policy generation unit is trained using the following steps: For the current execution step, determine the target value of the value function subnetwork corresponding to the current execution step. The target value is used to indicate the benefit from the current execution step to the termination execution step during the approach and landing process of the test aircraft in this round of training. The temporal difference target value is determined based on the training samples at the current execution step, the test sensing base station execution strategy and the reward value, and the training samples at the next execution step; The error loss is determined based on the target value of the value function subnetwork corresponding to the current execution step and the temporal difference target value; The policy generation unit is trained by minimizing the error loss.

[0084] It can be understood that each training round corresponds to one multi-vehicle approach and landing process, and each execution step in each training round corresponds to a different time point. For the current execution step, a value function can be constructed to determine the target value of the value function subnetwork corresponding to the current execution step. Specifically, the value function can be expressed by the following formula (22): (twenty two) in, This represents the target value of the Critic network at time t. Let T represent the discount factor, and T represent the end time of this training round. Represents the state space at time t. Represents the action space at time t. Indicates time The reward value.

[0085] During the policy update process, this embodiment employs a Temporal Difference (TD) objective function to train the Critic network. Specifically, at each decision time, a TD target value is determined based on the training samples at the current execution step, the test sensing base station's execution policy and the reward value, and the training samples at the next execution step. This TD target value comprehensively considers the immediate reward at the current moment and the potential future reward at the next moment under the test sensing base station's execution policy, thereby achieving an approximate estimate of long-term returns without requiring complete trajectory replay. An error loss is determined based on the target value of the value function sub-network corresponding to the current execution step and the TD target value. This error loss indicates the error between the Critic network output and the TD target value.

[0086] For example, the error loss can be determined by the following formula (23): (twenty three) in, Indicates error loss, This represents the target value of the Critic network at time t. This represents the TD target value at time t.

[0087] By minimizing error loss, the Critic network can gradually learn a more accurate action value function, providing a reliable basis for the parameter update guidance of the Actor network, thereby adjusting the parameters of the Actor network.

[0088] In this way, through the synergistic effect of the Actor-Critic structure, TD target value update and deterministic policy gradient, the reinforcement learning algorithm of this embodiment can continuously learn the approach guidance strategy of the sensing base station for eVTOL and the attitude control of the externally suspended take-off and landing platform in a complex urban low-altitude sensing environment. This reduces the aircraft's in-flight waiting time and improves the overall operational efficiency of the take-off and landing field while ensuring communication reliability and approach safety.

[0089] In some possible implementations, determining the temporal difference target value based on the training samples at the current execution step, the test sensing base station execution strategy and the reward value, and the training samples at the next execution step includes: Based on the training samples, the test sensing base station execution strategy, and the reward value at the current execution step, and the training samples at the next execution step, the predicted test sensing base station execution strategy for the next execution step is predicted. Based on the predicted test sensing base station execution strategy under the next execution step, predict the prediction value of the value function sub-network corresponding to the next execution step. The prediction value is used to indicate the benefit between the next execution step and the termination execution step during the approach and landing process of the test aircraft under this round of training. The temporal difference target value is determined based on the reward value at the current execution step and the prediction value of the value function subnetwork corresponding to the next execution step.

[0090] Specifically, the time-series difference target value can be determined by the following formula (24): (twenty four) in, This represents the TD target value at time t. This represents the reward value at time t. Indicates the discount factor. This represents the predictive value of the Critic network at time t+1. Represents the state space at time t+1. This represents the predicted action space at time t+1 (i.e., the predicted test sensor base station execution strategy for the next execution step).

[0091] In this way, by using the training samples, policy, and reward value at the current execution step, and the training samples at the next execution step, the policy at the next execution step is predicted and its predictive value of the Critic network is estimated. Then, the reward value at the current execution step and the predicted value are fused to generate a temporal difference target value. This allows the Critic network to accurately capture the overall change in revenue from the current moment to the end of landing during the training process. The Actor network continuously corrects its output accordingly, so that the trained policy prediction model can achieve stable and efficient landing with minimal alignment error, optimal communication capacity, and lowest energy consumption in high-density approach scenarios.

[0092] S103: According to the execution strategy of the sensing base station, control the sensing base station to communicate with the multiple aircraft, and control the sensing base station to adjust the multiple externally suspended take-off and landing platforms.

[0093] In this step, the communication power allocation ratio between the sensing base station and each aircraft, as included in the sensing base station execution strategy, can be used to control the communication between the sensing base station and the multiple aircraft; the horizontal rotation control amount of each externally suspended take-off and landing platform, as included in the sensing base station execution strategy, can be used to control the horizontal rotation mechanism to achieve left and right rotation of each externally suspended take-off and landing platform; the pitch control amount of each externally suspended take-off and landing platform, as included in the sensing base station execution strategy, can be used to control the pitch movement mechanism to achieve up and down movement of each externally suspended take-off and landing platform.

[0094] This embodiment of the disclosure sets up a sensing base station at the top of a multi-layered rotating vertical takeoff and landing (VTOL) field. The sensing base station can sense the aircraft, communicate with the aircraft, and adjust the externally suspended takeoff and landing platform. Through a unified communication and sensing node (i.e., the sensing base station), the approach trajectory of the aircraft is identified, located, predicted, and scheduled, transforming the takeoff and landing field from a passive parking point into a smart hub with airspace monitoring, approach management, and enhanced communication capabilities. The sensing base station can adjust its beam direction according to the platform's rotation status and the aircraft's position, achieving more stable link coverage and improving the communication reliability and scheduling coordination capabilities of the terminal airspace.

[0095] The multi-layered rotating vertical takeoff and landing field (UAM) in this embodiment does not rely on large areas of flat land, making UAM infrastructure feasible in urban centers where height is limited and land is scarce. By concentrating multiple platforms in a vertical space through a tower structure, high-density air traffic handling capacity can be achieved in cities at minimal cost. Parallel access and rotatable adaptive docking of multiple platforms significantly reduce hovering time in the airspace, lowering propulsion energy consumption, operating costs, and safety hazards caused by low-altitude congestion. Simultaneous reception of multiple platforms, automatic platform alignment, and reduced waiting time enable aircraft to quickly enter the final descent and landing phase, avoiding the energy consumption and airspace pressure accumulated during prolonged circling, with particularly significant advantages in high-density operating environments.

[0096] After executing the aforementioned sensing base station execution strategy, the step of acquiring aircraft radar point cloud data can be repeated. Based on the new aircraft radar point cloud data, a new sensing base station execution strategy is obtained through the strategy prediction model. In accordance with the new sensing base station execution strategy, the sensing base station is controlled to communicate with the multiple aircraft and to adjust the multiple externally suspended take-off and landing platforms until each aircraft lands on the corresponding externally suspended take-off and landing platform.

[0097] The aircraft approach and landing assistance method provided in this application collects aircraft radar point cloud data through a sensing base station during the approach and landing of multiple aircraft. The sensing base station is located at the top of a multi-layer rotating vertical take-off and landing field, which includes multiple externally suspended take-off and landing platforms. An execution strategy for the sensing base station is generated based on the aircraft radar point cloud data to control communication between the sensing base station and multiple aircraft, and to control the sensing base station to adjust the multiple externally suspended take-off and landing platforms. This multi-layer rotating vertical take-off and landing field helps reduce the land area occupied and increases the aircraft's carrying capacity. Compared to a single-layer or planar layout, it significantly improves land use efficiency. Furthermore, the sensing base station can adjust the multiple externally suspended take-off and landing platforms, allowing them to simultaneously align with the aircraft's attitude during approach and landing, enabling the aircraft to quickly enter the descent phase, reducing the time the aircraft spends hovering in the airspace, lowering propulsion energy consumption, operating costs, and safety hazards caused by low-altitude congestion, and improving the safety and efficiency of aircraft approach and landing.

[0098] Those skilled in the art will understand that, in the above-described method of the specific implementation, the order in which each step is written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0099] Corresponding to the aforementioned embodiments of the aircraft approach and landing assistance method, this application also provides embodiments of the aircraft approach and landing assistance device.

[0100] Please see Figure 4 This is a schematic diagram illustrating an aircraft approach and landing assistance device, as shown in an exemplary embodiment of this application. Figure 4 As shown in the illustration, the aircraft approach and landing assistance device 400 provided in this application embodiment includes: The data acquisition module 401 is used to acquire radar point cloud data of aircraft collected by the sensing base station during the approach and landing process of multiple aircraft; the sensing base station is set at the top of the multi-layer rotating vertical take-off and landing field, which includes multiple externally suspended take-off and landing platforms. The strategy prediction module 402 is used to input the radar point cloud data of the aircraft into a pre-trained strategy prediction model to obtain the sensor base station execution strategy output by the strategy prediction model. The sensor base station execution strategy is used to instruct the sensor base station to perform actions in the process of providing communication services and landing alignment services to the aircraft. The strategy execution module 403 is used to execute the strategy of the sensing base station, control the sensing base station to communicate with the multiple aircraft, and control the sensing base station to adjust the multiple externally suspended take-off and landing platforms.

[0101] In one optional implementation, the policy prediction model includes a radar point cloud feature extraction unit and a policy generation unit; The strategy prediction module 402 is specifically used for: The radar point cloud feature extraction unit extracts features from the aircraft radar point cloud data to obtain a radar point cloud feature vector. The aircraft radar point cloud data includes the aircraft position, the aircraft azimuth angle, the aircraft pitch angle, the relative distance between the sensing base station and the aircraft, the relative speed between the sensing base station and the aircraft, and the radar signal-to-noise ratio of the aircraft. The strategy generation unit performs strategy reasoning based on the radar point cloud feature vector to obtain the execution strategy of the sensing base station; the execution strategy of the sensing base station includes the communication power allocation ratio between the sensing base station and each aircraft, the horizontal rotation control amount of each externally suspended take-off and landing platform, and the pitch control amount of each externally suspended take-off and landing platform.

[0102] In an optional embodiment, the aircraft approach and landing assistance device 400 further includes a model training module 404, which is used to obtain the policy prediction model by training a reinforcement learning network. A state space representing the environmental information of the aircraft's approach and landing is constructed. Based on the state space, training samples are generated, including radar point cloud data of the test aircraft collected by the test sensing base station. The training samples are input into the reinforcement learning network to obtain the test sensory base station execution strategy output by the reinforcement learning network. When executing the test sensing base station execution strategy, determine the platform selection cost corresponding to each test aircraft choosing to land on an externally suspended take-off and landing platform, and determine the reward value corresponding to the test sensing base station execution strategy; Minimizing the platform selection cost is determined as the optimization objective. The reinforcement learning network is iteratively trained until the network meets the cutoff condition based on the reward value. The reinforcement learning network that meets the cutoff condition is then determined as the trained policy prediction model.

[0103] In one optional implementation, the model training module 404 is used to determine the platform selection cost corresponding to landing on an externally suspended take-off and landing platform for any of the test aircraft through the following steps: Based on the position of the test aircraft and the position of the externally suspended take-off and landing platform selected by the test aircraft for landing, the distance error between the test aircraft and the corresponding externally suspended take-off and landing platform is determined; Based on the azimuth angle of the test aircraft and the corresponding horizontal rotation angle of the externally suspended take-off and landing platform, the horizontal alignment error between the test aircraft and the corresponding externally suspended take-off and landing platform is determined. Based on the pitch angle of the test aircraft and the corresponding pitch angle of the externally suspended take-off and landing platform, determine the pitch direction alignment error between the test aircraft and the corresponding externally suspended take-off and landing platform; Based on the communication bandwidth between the test aircraft and the corresponding external take-off and landing platform and the signal-to-noise ratio of the communication link of the test aircraft, the communication channel capacity between the test aircraft and the corresponding external take-off and landing platform is determined. Based on the distance error, horizontal alignment error, pitch alignment error, and communication channel capacity between the test aircraft and the corresponding externally suspended take-off and landing platform, the platform selection cost corresponding to the test aircraft choosing the externally suspended take-off and landing platform for landing is determined.

[0104] In one optional implementation, the model training module 404 is used to determine whether the reinforcement learning network meets the cutoff condition through the following steps: Based on the communication channel capacity, distance error, horizontal alignment error, pitch alignment error, horizontal rotation control amount and pitch control amount of each externally suspended take-off and landing platform between the test aircraft and the corresponding externally suspended take-off and landing platform, the reward value corresponding to the execution strategy of the test sensing base station is determined. When the reward value tends to stabilize, the reinforcement learning network is determined to meet the cutoff condition.

[0105] In one optional implementation, the policy generation unit includes a policy sub-network and a value function sub-network. The policy sub-network is used to output the execution policy of the test sensing base station, and the value function sub-network is used to evaluate the value of the execution policy of the test sensing base station and generate a reward value corresponding to the execution policy of the test sensing base station. Each training round includes multiple consecutive execution steps, and the corresponding execution policy and reward value of the test sensing base station are obtained based on the training samples at each execution step. The model training module 404 is used to train the policy generation unit through the following steps: For the current execution step, determine the target value of the value function subnetwork corresponding to the current execution step. The target value is used to indicate the benefit from the current execution step to the termination execution step during the approach and landing process of the test aircraft in this round of training. The temporal difference target value is determined based on the training samples at the current execution step, the test sensing base station execution strategy and the reward value, and the training samples at the next execution step; The error loss is determined based on the target value of the value function subnetwork corresponding to the current execution step and the temporal difference target value; The policy generation unit is trained by minimizing the error loss.

[0106] In an optional implementation, the model training module 404, when determining the temporal difference target value based on the training samples at the current execution step, the test sensing base station execution strategy, the reward value, and the training samples at the next execution step, is specifically used for: Based on the training samples, the test sensing base station execution strategy, and the reward value at the current execution step, and the training samples at the next execution step, the predicted test sensing base station execution strategy for the next execution step is predicted. Based on the predicted test sensing base station execution strategy under the next execution step, predict the prediction value of the value function sub-network corresponding to the next execution step. The prediction value is used to indicate the benefit between the next execution step and the termination execution step during the approach and landing process of the test aircraft under this round of training. The temporal difference target value is determined based on the reward value at the current execution step and the prediction value of the value function subnetwork corresponding to the next execution step.

[0107] The specific implementation process of the functions and roles of each module in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0108] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0109] Based on the same technical concept, this application also provides a computer device 500, referring to... Figure 5 The diagram shown is a schematic representation of the structure of a computer device according to an exemplary embodiment of this application, comprising: The processor 510, memory 520, and bus 530 are included. The memory 520 is used to store execution instructions and includes main memory 521 and external memory 522. The main memory 521, also known as internal memory, is used to temporarily store the operation data in the processor 510 and the data exchanged with external memory 522 such as hard disk. The processor 510 exchanges data with external memory 522 through main memory 521.

[0110] In this embodiment, the memory 520 is specifically used to store application code that executes the solution of this application, and its execution is controlled by the processor 510. That is, when the electronic device 500 is running, the processor 510 communicates with the memory 520 through the bus 530, or the processor 510 communicates with the memory 520 through other means, so that the processor 510 executes the application code stored in the memory 520, and then executes the steps of the aircraft approach and landing assistance method described in any of the foregoing embodiments.

[0111] The memory 520 may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.

[0112] Processor 510 may be an integrated circuit chip with signal processing capabilities. The aforementioned processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor.

[0113] It is understood that the structures illustrated in the embodiments of this application do not constitute a specific limitation on the electronic device 500. In other embodiments of this application, the electronic device 500 may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0114] This disclosure also provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the aircraft approach and landing assistance method described in the above-described method embodiments. The storage medium may be a volatile or non-volatile computer-readable storage medium.

[0115] This disclosure also provides a computer program product storing a computer program. When the computer program is run by a processor, it executes the steps of the aircraft approach and landing assistance method provided in any of the above embodiments of this disclosure. For details, please refer to the above method embodiments, which will not be repeated here.

[0116] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium, which can be a volatile or non-volatile computer-readable storage medium. In another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0117] Furthermore, embodiments of the subject matter and functional operation described in this specification can be implemented in the following ways: digital electronic circuits, tangibly embodied computer software or firmware, computer hardware including the structures disclosed in this specification and their structural equivalents, or combinations thereof. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions encoded on a tangible, non-transitory program carrier for execution by a data processing apparatus or for controlling the operation of a data processing apparatus. Alternatively or additionally, program instructions may be encoded on artificially generated propagation signals, such as machine-generated electrical, optical, or electromagnetic signals, which are generated to encode information and transmit it to a suitable receiving device for execution by the data processing apparatus. The computer storage medium may be a machine-readable storage device, a machine-readable storage substrate, a random or serial access memory device, or combinations thereof.

[0118] The processing and logic flow described in this specification can be executed by one or more programmable computers that execute one or more computer programs to perform corresponding functions by operating on input data and generating output. The processing and logic flow can also be executed by dedicated logic circuitry—such as FPGAs (Field-Programmable Gate Arrays) or ASICs (Application-Specific Integrated Circuits), and the device can also be implemented as dedicated logic circuitry.

[0119] Suitable computers for executing computer programs include, for example, general-purpose and / or special-purpose microprocessors, or any other type of central processing unit. Typically, the central processing unit receives instructions and data from read-only memory and / or random access memory. The basic components of a computer include a central processing unit for implementing or executing instructions and one or more memory devices for storing instructions and data. Typically, a computer will also include one or more mass storage devices for storing data, such as disks, magneto-optical disks, or optical disks, or the computer will be operatively coupled to such mass storage devices to receive data from or transfer data to them, or both. However, a computer is not required to have such devices. Furthermore, a computer can be embedded in another device, such as a mobile phone, a personal digital assistant (PDA), a mobile audio or video player, a game console, a global positioning system (GPS) receiver, or a portable storage device such as a universal serial bus (USB) flash drive, to name a few.

[0120] Computer-readable media suitable for storing computer program instructions and data include all forms of non-volatile memory, media, and memory devices, such as semiconductor memory devices (e.g., EPROM, EEPROM, and flash memory devices), magnetic disks (e.g., internal hard disks or removable disks), magneto-optical disks, and CD-ROM and DVD-ROM disks. Processors and memory may be supplemented by or incorporated into dedicated logic circuitry.

[0121] While this specification contains numerous specific implementation details, these should not be construed as limiting the scope of any invention or the scope of the claims, but rather are primarily intended to describe features of specific embodiments of a particular invention. Certain features described in the various embodiments herein may also be implemented in combination in a single embodiment. Conversely, various features described in a single embodiment may also be implemented separately in various embodiments or in any suitable sub-combination. Furthermore, while features may function in certain combinations as described above and even initially claimed in this way, one or more features from a claimed combination may be removed from that combination in some cases, and a claimed combination may refer to a sub-combination or a variation thereof.

[0122] Similarly, although the operations are depicted in a specific order in the accompanying drawings, this should not be construed as requiring these operations to be performed in the specific order shown or sequentially, or requiring all illustrated operations to be performed to achieve the desired result. In some cases, multitasking and parallel processing may be advantageous. Furthermore, the separation of various system modules and components in the above embodiments should not be construed as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.

[0123] Thus, specific embodiments of the subject matter have been described. Other embodiments are within the scope of the appended claims. In some cases, the actions recited in the claims may be performed in a different order and still achieve the desired result. Furthermore, the processes depicted in the drawings are not necessarily shown in a specific order or sequence to achieve the desired result. In some implementations, multitasking and parallel processing may be advantageous.

[0124] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for assisting aircraft approach and landing, characterized in that, The method includes: During the approach and landing of multiple aircraft, radar point cloud data of the aircraft collected by the sensing base station is acquired; the sensing base station is set at the top of the multi-layer rotating vertical take-off and landing field, which includes multiple externally suspended take-off and landing platforms. The radar point cloud data of the aircraft is input into a pre-trained strategy prediction model to obtain the sensor base station execution strategy output by the strategy prediction model. The sensor base station execution strategy is used to instruct the sensor base station to perform actions in the process of providing communication services and landing alignment services to the aircraft. According to the execution strategy of the sensing base station, control the sensing base station to communicate with the multiple aircraft, and control the sensing base station to adjust the multiple externally suspended take-off and landing platforms; The strategy prediction model includes a radar point cloud feature extraction unit and a strategy generation unit. The step of inputting the aircraft radar point cloud data into a pre-trained policy prediction model to obtain the sensing base station execution policy output by the policy prediction model includes: The radar point cloud feature extraction unit extracts features from the aircraft radar point cloud data to obtain a radar point cloud feature vector. The aircraft radar point cloud data includes the aircraft position, the aircraft azimuth angle, the aircraft pitch angle, the relative distance between the sensing base station and the aircraft, the relative speed between the sensing base station and the aircraft, and the radar signal-to-noise ratio of the aircraft. The strategy generation unit performs strategy reasoning based on the radar point cloud feature vector to obtain the execution strategy of the sensing base station; the execution strategy of the sensing base station includes the communication power allocation ratio between the sensing base station and each aircraft, the horizontal rotation control amount of each externally suspended take-off and landing platform, and the pitch control amount of each externally suspended take-off and landing platform. The policy prediction model is obtained by training a reinforcement learning network. A state space representing the environmental information of the aircraft's approach and landing is constructed. Based on the state space, training samples are generated, including radar point cloud data of the test aircraft collected by the test sensing base station. The training samples are input into the reinforcement learning network to obtain the test sensory base station execution strategy output by the reinforcement learning network. When executing the test sensing base station execution strategy, determine the platform selection cost corresponding to each test aircraft choosing to land on an externally suspended take-off and landing platform, and determine the reward value corresponding to the test sensing base station execution strategy; Minimize the platform selection cost as the optimization objective, iteratively train the reinforcement learning network until the reinforcement learning network meets the cutoff condition based on the reward value, and the reinforcement learning network that meets the cutoff condition is determined as the trained policy prediction model. The platform selection cost for any of the test aircraft to land on an externally suspended take-off and landing platform is determined through the following steps: The weighted summation of the distance error, horizontal alignment error, and pitch alignment error between the test aircraft and the corresponding externally suspended take-off and landing platform with their respective weighting coefficients is determined. Determine the product of the communication channel capacity between the test aircraft and the corresponding external take-off and landing platform and the corresponding weighting coefficient; The result obtained by subtracting the product from the weighted summation result is determined as the platform selection cost.

2. The method according to claim 1, characterized in that, Before determining the weighted summation of the distance error, horizontal alignment error, and pitch alignment error between the test aircraft and the corresponding externally suspended take-off and landing platform with their respective weighting coefficients, the process also includes: Based on the position of the test aircraft and the position of the externally suspended take-off and landing platform selected by the test aircraft for landing, the distance error between the test aircraft and the corresponding externally suspended take-off and landing platform is determined; Based on the azimuth angle of the test aircraft and the corresponding horizontal rotation angle of the externally suspended take-off and landing platform, the horizontal alignment error between the test aircraft and the corresponding externally suspended take-off and landing platform is determined. Based on the pitch angle of the test aircraft and the corresponding pitch angle of the externally suspended take-off and landing platform, determine the pitch direction alignment error between the test aircraft and the corresponding externally suspended take-off and landing platform; Based on the communication bandwidth between the test aircraft and the corresponding externally suspended take-off and landing platform, and the signal-to-noise ratio of the communication link of the test aircraft, the communication channel capacity between the test aircraft and the corresponding externally suspended take-off and landing platform is determined.

3. The method according to claim 1, characterized in that, The following steps are used to determine whether the reinforcement learning network meets the cutoff condition: Based on the communication channel capacity, distance error, horizontal alignment error, pitch alignment error, horizontal rotation control amount and pitch control amount of each externally suspended take-off and landing platform between the test aircraft and the corresponding externally suspended take-off and landing platform, the reward value corresponding to the execution strategy of the test sensing base station is determined. When the reward value tends to stabilize, the reinforcement learning network is determined to meet the cutoff condition.

4. The method according to claim 1, characterized in that, The strategy generation unit includes a strategy sub-network and a value function sub-network. The strategy sub-network is used to output the execution strategy of the test sensing base station, and the value function sub-network is used to evaluate the value of the execution strategy of the test sensing base station and generate the reward value corresponding to the execution strategy of the test sensing base station. Each round of training includes multiple consecutive execution steps, and the corresponding execution strategy and reward value of the test sensing base station are obtained based on the training samples at each execution step. The policy generation unit is trained using the following steps: For the current execution step, determine the target value of the value function subnetwork corresponding to the current execution step. The target value is used to indicate the benefit from the current execution step to the termination execution step during the approach and landing process of the test aircraft in this round of training. The temporal difference target value is determined based on the training samples at the current execution step, the test sensing base station execution strategy and the reward value, and the training samples at the next execution step; The error loss is determined based on the target value of the value function subnetwork corresponding to the current execution step and the temporal difference target value; The policy generation unit is trained by minimizing the error loss.

5. The method according to claim 4, characterized in that, The step of determining the temporal difference target value based on the training samples at the current execution step, the test sensing base station execution strategy, the reward value, and the training samples at the next execution step includes: Based on the training samples, the test sensing base station execution strategy, and the reward value at the current execution step, and the training samples at the next execution step, the predicted test sensing base station execution strategy for the next execution step is predicted. Based on the predicted test sensing base station execution strategy under the next execution step, predict the prediction value of the value function sub-network corresponding to the next execution step. The prediction value is used to indicate the benefit between the next execution step and the termination execution step during the approach and landing process of the test aircraft under this round of training. The temporal difference target value is determined based on the reward value at the current execution step and the prediction value of the value function subnetwork corresponding to the next execution step.

6. An aircraft approach and landing assistance device, characterized in that, The device includes: The data acquisition module is used to acquire radar point cloud data of aircraft collected by the sensing base station during the approach and landing of multiple aircraft; the sensing base station is set at the top of the multi-layer rotating vertical take-off and landing field, which includes multiple externally suspended take-off and landing platforms. The strategy prediction module is used to input the radar point cloud data of the aircraft into a pre-trained strategy prediction model to obtain the sensor base station execution strategy output by the strategy prediction model. The sensor base station execution strategy is used to instruct the sensor base station to perform actions in the process of providing communication services and landing alignment services to the aircraft. The strategy execution module is used to execute the strategy according to the sensing base station, control the sensing base station to communicate with the multiple aircraft, and control the sensing base station to adjust the multiple externally suspended take-off and landing platforms; The strategy prediction model includes a radar point cloud feature extraction unit and a strategy generation unit. The strategy prediction module is specifically used for: The radar point cloud feature extraction unit extracts features from the aircraft radar point cloud data to obtain a radar point cloud feature vector. The aircraft radar point cloud data includes the aircraft position, the aircraft azimuth angle, the aircraft pitch angle, the relative distance between the sensing base station and the aircraft, the relative speed between the sensing base station and the aircraft, and the radar signal-to-noise ratio of the aircraft. The strategy generation unit performs strategy reasoning based on the radar point cloud feature vector to obtain the execution strategy of the sensing base station; the execution strategy of the sensing base station includes the communication power allocation ratio between the sensing base station and each aircraft, the horizontal rotation control amount of each externally suspended take-off and landing platform, and the pitch control amount of each externally suspended take-off and landing platform. The model training module is used to obtain the policy prediction model by training the reinforcement learning network. A state space representing the environmental information of the aircraft's approach and landing is constructed. Based on the state space, training samples are generated, including radar point cloud data of the test aircraft collected by the test sensing base station. The training samples are input into the reinforcement learning network to obtain the test sensory base station execution strategy output by the reinforcement learning network. When executing the test sensing base station execution strategy, determine the platform selection cost corresponding to each test aircraft choosing to land on an externally suspended take-off and landing platform, and determine the reward value corresponding to the test sensing base station execution strategy; Minimize the platform selection cost as the optimization objective, iteratively train the reinforcement learning network until the reinforcement learning network meets the cutoff condition based on the reward value, and the reinforcement learning network that meets the cutoff condition is determined as the trained policy prediction model. The model training module is used to determine the platform selection cost corresponding to landing on an externally suspended take-off and landing platform for any of the test aircraft through the following steps: The weighted summation of the distance error, horizontal alignment error, and pitch alignment error between the test aircraft and the corresponding externally suspended take-off and landing platform with their respective weighting coefficients is determined. Determine the product of the communication channel capacity between the test aircraft and the corresponding external take-off and landing platform and the corresponding weighting coefficient; The result obtained by subtracting the product from the weighted summation result is determined as the platform selection cost.

7. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the aircraft approach and landing assistance method according to any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the aircraft approach and landing assistance method according to any one of claims 1 to 6.

Images