Low-altitude aircraft real-time monitoring and guiding method based on 5g-a

Abstract

The application relates to the technical field of low-altitude aircraft real-time monitoring and guiding methods based on 5G-A, and specifically discloses a low-altitude aircraft real-time monitoring and guiding method based on 5G-A, aiming to solve the problems of a large number of low-altitude airspace perception blind areas, extensive dynamic management and lagging guiding response. The method comprises the following steps: constructing a cloud-network-edge-end collaborative architecture; performing multi-level fusion perception; implementing dynamic airspace cellular management; performing conflict prediction and flight path pre-optimization; and issuing a structured guiding instruction. Through the above technical scheme, the application realizes global continuous perception, fine airspace management and millisecond-level closed-loop guiding, and significantly improves low-altitude flight safety and airspace utilization efficiency.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of wireless communication and air traffic management technology. Specifically, it relates to a method for real-time status perception, risk assessment and dynamic trajectory guidance of various aircraft in urban and near-ground low-altitude areas by utilizing the enhanced network capabilities of fifth-generation mobile communication technology. Background Technology

[0002] With the rapid development of the low-altitude economy, the number of new aircraft such as drones, eVTOLs, and helicopters in urban and near-ground low-altitude airspace is growing exponentially, and the operational scenarios are becoming increasingly complex and dense. Traditional low-altitude monitoring and guidance systems mainly rely on radar, ADS-B broadcast surveillance, or optical vision systems, which generally suffer from inherent defects such as discontinuous perception coverage, limited communication link bandwidth, excessively high response latency, and a lack of multi-source collaborative decision-making mechanisms. Especially in urban canyons, dense building clusters, or areas with complex electromagnetic interference, the above methods are difficult to achieve stable tracking of small, low-speed, and non-cooperative targets, and cannot obtain high-dimensional semantic information such as the aircraft's attitude angle, dynamic state, and mission intent, which seriously restricts the safety capacity and operational efficiency of low-altitude airspace.

[0003] Among them, 5G-Advanced technology, with its native capabilities such as integrated sensing, network slicing, edge intelligence, and centimeter-level positioning, provides the technological possibility for building a new generation of low-altitude intelligent control infrastructure. However, existing low-altitude management solutions based on 5G-A still remain at the level of single-point perception or static compliance verification. For example, they only use base station echo signals for position comparison to identify illegal flights, lacking in-depth analysis of the aircraft's health status and dynamic intentions; while multi-base station collaboration improves trajectory tracking accuracy, its function stops at anomaly alarms, failing to establish a closed-loop control link from risk prediction to trajectory replanning. More importantly, these solutions generally ignore the differentiated QoS guarantee capabilities provided by 5G-A network slicing, failing to ensure deterministic transmission of emergency guidance commands under ultra-low latency and ultra-high reliability constraints, and failing to deeply integrate sensing data, communication resources, and control strategies under the "cloud-network-edge-device" architecture, resulting in a fragmented state of "strong monitoring, weak guidance" and "data available, but no action taken."

[0004] Therefore, there is an urgent need for a real-time monitoring and guidance method for low-altitude aircraft based on 5G-A, which can deeply integrate sensor signals, digital twins and AI decision-making to achieve a complete closed loop in dynamic airspace, from continuous perception across the entire domain and millisecond-level risk prediction to the issuance of structured flight path instructions, thereby breaking through the multiple bottlenecks of existing technologies in terms of perception dimension, communication support, guidance intelligence and system coordination. Summary of the Invention

[0005] The purpose of this invention is to provide a real-time monitoring and guidance method for low-altitude aircraft based on 5G-A, which can effectively solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0007] The 5G-A-based method for real-time monitoring and guidance of low-altitude aircraft includes the following specific steps:

[0008] Step S1: Construct a cloud-network-edge-device collaborative architecture, deploy global airspace management strategies and AI training models on the central control cloud platform, deploy edge computing nodes and run a low-altitude digital twin engine on the 5G-A ground base station side, integrate a communication and sensing integrated module on the aircraft terminal, and achieve highly reliable and low-latency data interaction between layers through low-altitude dedicated network slicing.

[0009] Step S2: Perform multi-level fusion perception, use multiple 5G-A base stations to perform passive three-dimensional positioning of the target based on the signal arrival time difference and arrival angle, analyze the channel state information of the aircraft's uplink signal to monitor its flight status non-intrusively, and exchange local obstacle perception data through the side links between aircraft. The base station perception data, aircraft reported data and external environment data are spatiotemporally aligned and fused in the edge digital twin to generate a high-dimensional flight situation.

[0010] Step S3: Implement dynamic airspace cell management, divide the low-altitude airspace into non-uniform virtual airspace cells, dynamically adjust the geometric boundaries and passage rules of each cell according to real-time traffic density, meteorological conditions and the status of sensitive ground areas, and broadcast the updated airspace strategy to relevant aircraft through 5G-A network slicing.

[0011] Step S4: Perform conflict prediction and trajectory pre-optimization. Based on the real-time situation in the edge digital twin, use a lightweight AI model to extrapolate traffic flow in a specific time period in the future, identify potential conflicts or violation risks, and combine the mission type, performance constraints and remaining energy of the affected aircraft to generate one or more four-dimensional collaborative escape trajectories in the digital twin environment.

[0012] Step S5: Issue structured guidance instructions, converting the optimized track into structured instructions containing time, speed, altitude and waypoint coordinates. These instructions are then sent to the relevant aircraft flight control system via the 5G-A control channel with end-to-end latency less than a preset threshold and reliability not lower than a preset reliability, ensuring quality of service. The flight control system can then directly parse and execute the instructions or execute them after operator confirmation.

[0013] Furthermore, in step S1, the central control cloud platform is responsible for macro-airspace planning, global AI model training, and data docking with the civil aviation supervision system. Edge computing nodes are distributed and deployed on each 5G-A base station side, running a dynamic three-dimensional airspace image with predetermined accuracy and predetermined update cycle. The communication and sensing integrated module of the aircraft terminal supports the transmission and reception of modulated sensing signals for detecting surrounding obstacles.

[0014] Furthermore, in step S2, the accuracy of signal arrival time difference positioning is better than a preset accuracy threshold, the arrival angle resolution reaches a preset angle resolution, and the channel state information analysis constructs the aircraft vibration intensity index by extracting the time delay spread and Doppler frequency shift characteristics of the multipath components. ,in For the first The time delay deviation of each multipath component, For its Doppler frequency shift, The total number of effective multipaths.

[0015] Furthermore, in step S2, the inter-vehicle side walkway adopts a distributed scheduling mechanism based on a resource pool. The sensing data exchange cycle is no greater than a preset time threshold. The exchanged content includes obstacle point cloud coordinates, relative distances, and confidence weights. The weight calculation formula is as follows: ,in Distance to the obstacle The relative azimuth angle. , , For preset coefficients and .

[0016] Furthermore, in step S2, the multi-source data fusion employs a joint correction mechanism of Kalman filtering and deep learning. The state vector includes position, velocity, attitude angle, and intention label. The observation noise covariance matrix R is dynamically adjusted according to the signal-to-noise ratio. ,in , This is an empirical constant.

[0017] Furthermore, in step S3, the initial division of the spatial domain cells is based on a hexagonal grid with a side length of... ,in The maximum permitted flight speed in this area. For safe intervals; local traffic density When the density exceeds a preset threshold, the cell automatically divides into multiple daughter cells.

[0018] Furthermore, in step S4, the conflict prediction model adopts a graph neural network architecture. Node features include aircraft position, velocity vector, and mission priority; edge features are relative distance and proximity rate; and the conflict determination condition is... and ,in Predict the minimum spacing at time t. For safe distance threshold, This is the separation speed threshold.

[0019] Furthermore, the objective function for four-dimensional trajectory optimization in step S4 is: ,in For additional energy consumption, Delays caused by the task For the first The acceleration of the flight path, , , For normalized weight coefficients, satisfying .

[0020] Furthermore, the structured instruction format in step S5 follows an extension of the ISO 21384-3 standard, including timestamp T, target velocity V, target altitude H, and three-dimensional waypoint coordinates. The instruction parsing delay does not exceed the preset delay threshold, and the flight control system execution response time is less than the preset response time threshold.

[0021] Furthermore, the uplink bandwidth of the low-altitude dedicated network slice is configured to be no less than a preset bandwidth threshold, and the downlink control channel adopts repetitive coding and polar code concatenation, with a maximum retransmission count of a preset number, ensuring that the end-to-end delay is less than a preset delay threshold within the predetermined coverage area.

[0022] Furthermore, the edge digital twin engine updates the spatial domain image once at a predetermined update cycle. The update process includes data alignment, anomaly removal, and state interpolation, wherein the time alignment error is less than a preset time error threshold, and the spatial registration error is less than a preset spatial error threshold.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] 1. Significantly Enhanced Continuous Sensing Capabilities Across the Entire Domain: By simultaneously using 5G-A base stations and aircraft terminals as communication and sensing nodes, deep integration of "infrastructure as sensor" is achieved. The network side utilizes multiple base stations... Achieving passive high-precision positioning, the terminal side constructs an airborne self-organizing perception network through side links, effectively compensating for perception blind spots in complex scenarios such as urban canyons. Simultaneously, by non-intrusively acquiring aircraft health status through CSI signal analysis, the perception dimension is expanded from single-location information to a high-dimensional semantic space including attitude, intent, and dynamic state, significantly improving the detection success rate of small, low-speed, and non-cooperative targets.

[0025] 2. Dynamic airspace management achieves refinement and agility: The proposed "dynamic airspace cell" concept breaks through the limitations of traditional static airspace division. The geometry and traffic rules of the cells can be dynamically adjusted according to real-time traffic flow, weather, and the status of sensitive ground areas, and strategies can be rapidly deployed through 5G-A network slicing. This mechanism significantly improves airspace utilization efficiency, supporting aircraft densities far exceeding the limits of existing systems during peak hours.

[0026] 3. Intelligent guidance enables millisecond-level closed-loop control: This invention is the first to achieve a complete closed loop from "perception-prediction-decision-control" at the edge. Conflict prediction based on digital twins has sufficient lead time, and trajectory optimization comprehensively considers mission, performance, and energy constraints. The generated structured commands can be directly executed by the flight control system. Real-world testing shows that the average intervention latency in emergency obstacle avoidance scenarios is significantly reduced, substantially lower than traditional warning modes, effectively avoiding the vast majority of potential conflicts.

[0027] 4. The system architecture boasts high scalability and smooth evolution capabilities: The "cloud-edge-device" collaborative architecture achieves optimal distribution of computing power: the cloud handles global policy and model training, the edge processes latency-sensitive real-time inference, and the terminal executes lightweight sensing and control. This design not only meets the capability boundaries of the current 5G-A network but also reserves interfaces for future 6G integrated sensing and computing evolution. Reusing existing cellular infrastructure deployment methods significantly reduces deployment costs per unit area, making large-scale commercialization feasible. Attached Figure Description

[0028] Figure 1 A schematic diagram of the core steps of the real-time monitoring and guidance method for low-altitude aircraft based on 5G-A of this invention.

[0029] Figure 2 This is a schematic diagram of the overall technical solution architecture of the 5G-A-based real-time monitoring and guidance method for low-altitude aircraft proposed in this invention.

[0030] Figure 3 This is a schematic diagram of the core principle framework of multi-level fusion perception and high-dimensional flight situation generation in this invention;

[0031] Figure 4 This is a logical flowchart of the dynamic airspace cellular management and strategy broadcasting in this invention;

[0032] Figure 5 This is a logical flowchart of the conflict prediction and four-dimensional trajectory pre-optimization based on edge digital twins in this invention.

[0033] Figure 6 This is a schematic diagram of the multi-level interaction relationship and data flow of the structured guidance instruction issuance and execution under the cloud-network-edge-device collaborative architecture in this invention; Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0035] Example 1

[0036] In the aforementioned 5G-A-based real-time monitoring and guidance method for low-altitude aircraft, step S1, which constructs a cloud-network-edge-device collaborative architecture, specifically includes the following operations: A central control cloud platform is deployed in a national or regional data center, running a global airspace management strategy engine and an AI training model cluster. This platform connects bidirectionally with the Civil Aviation Administration's UOM system and the public security low-altitude security platform via standardized API interfaces, receiving macro-level airspace opening instructions, no-fly zone updates, and emergency response commands, and simultaneously uploading anonymized flight activity statistics. The central control cloud platform adopts a distributed microservice architecture, including an airspace planning service module, an AI model training service module, a data lake storage module, and a cross-domain collaborative scheduling module. The AI ​​model training service module periodically collects local simulation logs, conflict event samples, and trajectory optimization results from each edge computing node, trains a lightweight graph neural network model using a federated learning mechanism, and encrypts and distributes the updated model parameters to each edge computing node.

[0037] The edge computing nodes deployed on the 5G-A terrestrial base station side adopt general-purpose server hardware, configured with at least 32-core CPUs, 128GB of memory, and dedicated AI acceleration cards. They are directly connected to the base station baseband unit via a PCIe 4.0 interface to ensure low-latency access to raw sensing data. Each edge computing node runs a low-altitude digital twin engine, which builds a dynamic three-dimensional airspace mirror with an update cycle of 100 milliseconds, achieving horizontal accuracy. Vertical direction Centimeters. The digital twin engine is internally divided into four functional sub-modules: a multi-source data access module, a spatiotemporal alignment module, a state fusion module, and a visualization output module. The multi-source data access module obtains data from the base station via the gRPC protocol. The raw measurements are received from the aircraft terminal via the MQTT protocol using ADS-BLike messages. Simultaneously, real-time wind speed and visibility data are obtained from the meteorological bureau's API via HTTPS, and 3D building models and geofences for sensitive areas are acquired from the city's GIS platform. The spatiotemporal alignment module employs a joint algorithm based on timestamp interpolation and spatial coordinate transformation to uniformly map heterogeneous data to the same spatiotemporal reference in the WGS-84 coordinate system. The time alignment error is controlled within ±2 milliseconds, and the spatial registration error is less than ±5 centimeters.

[0038] The integrated communication and sensing module at the aircraft's terminal layer employs a 5G-A chipset compliant with the 3GPP Release 18 standard, supporting integrated sensing waveform transmission and reception in the Sub-6GHz band. This module features a dual-channel RF front-end: one channel for regular communication data transmission and reception, and the other dedicated to transmitting chirp-modulated sensing signals with a bandwidth of 100MHz and a pulse repetition frequency of 1kHz. After directional transmission via the airborne antenna array, the echoes are received by the ISAC modules of the same module or neighboring aircraft. Matched filtering and FFT processing are used to extract obstacle distance, velocity, and orientation information. An embedded processor within the module runs a lightweight point cloud clustering algorithm, converting the raw echo data into a structured obstacle list, including obstacle center coordinates, size estimates, and motion vectors. All sensing data is reported to the edge computing node via the 5G-A uplink at intervals no greater than 50 milliseconds, and simultaneously broadcast to other aircraft within a 1km radius via the side link at intervals no greater than 100 milliseconds.

[0039] The low-altitude dedicated network slice is created collaboratively by the network slice selection and session management functions in the operator's core network, assigning an independent S-NSSAI identifier to low-altitude services. This slice is configured with a dedicated PRB resource pool on the Radio Access Network (RAN) side, SRv6 path assurance is enabled on the transport network side, and a dedicated UPF instance is deployed on the core network side. The slice's quality of service parameters are strictly set as follows: uplink bandwidth no less than 50 Mbps / aircraft, downlink control channel end-to-end latency less than 10 milliseconds, and reliability no less than 99.999%. To achieve these targets, the downlink control channel adopts a channel coding scheme of repetitive coding and polar code concatenation, with a maximum retransmission count of 3, and the HARQ feedback window compressed to 2 milliseconds.

[0040] In the above method, step S2 performs multi-level fusion sensing, specifically as follows: First, in the network-side active and passive sensing phase, at least three geographically distributed 5G-A ground base stations synchronously receive uplink reference signals (SRS) from the same aircraft. The radio frequency unit of each base station records the precise timestamp and incident angle of the signal arrival, with a timestamp accuracy better than 1 nanosecond and an angle resolution better than 0.5 degrees. These raw measurements are aggregated via the Xn interface to the edge computing node connected to the main base station serving the aircraft. Within the edge computing node, the TDOA positioning module uses the least squares method to solve the overdetermined equations, converting the time difference measurements from multiple base stations into three-dimensional spatial coordinates; the AOA positioning module uses beamforming weights to infer the target's azimuth. The results of both are fused through a weighted average, with the weights dynamically adjusted according to the signal-to-noise ratio of each base station. The root mean square error (RMSE) of the final positioning result is better than 30 centimeters in urban environments.

[0041] Meanwhile, the Channel State Information (CSI) analysis module continuously monitors the multipath component characteristics of the aircraft's uplink signal. The base station extracts the delay spread of each effective multipath by analyzing the phase and amplitude variations of the OFDM subcarriers. With Doppler frequency shift The system defines the aircraft vibration intensity index ( ). )for:

[0042]

[0043] in The effective total number of multipath components is typically selected based on a signal-to-noise ratio (SNR) greater than 15 dB. When the VIBI value exceeds a preset threshold, the system determines that the aircraft is in an abnormal vibration state, which may be caused by mechanical failure or strong airflow disturbance. This status label is then attached to the aircraft's digital twin attributes.

[0044] Secondly, in the terminal-side collaborative sensing phase, decentralized communication is established between aircraft via the 5G-A sidelink. The sidelink employs a distributed scheduling mechanism based on a resource pool, where each aircraft autonomously selects transmission resources from a predefined time-frequency resource pool, avoiding the overhead of centralized scheduling. The sensing data exchange period is set to 100 milliseconds, and the exchanged content includes obstacle point cloud coordinates (a local coordinate system with the aircraft itself as the origin), relative distance d, and confidence weight W. The formula for calculating weight W is:

[0045]

[0046] in Distance to obstacles (unit: meters) The azimuth angle of the obstacle relative to the aircraft's heading (unit: radians). , , The preset coefficients satisfy... Typical values ​​are , This weighting takes into account both distance attenuation and direction sensitivity, resulting in higher confidence for obstacles ahead. The received neighboring aircraft perception data undergoes local coordinate transformation to be unified into the global WGS-84 coordinate system before being uploaded to the edge computing node via uplink.

[0047] Finally, in the multi-source data fusion stage, the state fusion module of the edge digital twin engine adopts a joint correction mechanism of Kalman filtering and deep learning. (State vector) Defined as The first six terms are position and velocity, the middle three are Euler attitude angles, and the last term is the intention label (such as discrete values ​​like "logistics delivery" or "emergency rescue"). Observation vector By base station The system comprises the positioning results, the GNSS position reported by the aircraft, and the center point of the airborne visual inspection frame. The observation noise covariance matrix R_k is dynamically adjusted based on the real-time signal-to-noise ratio of each data source, and its expression is: ,in , It is an empirical constant (typically 0.5 meters). for Signal-to-noise ratio (SNR) of the orientation positioning signal (unit: dB). The Kalman filter's prediction step uses a constant acceleration model, while the update step introduces a lightweight LSTM network that takes the state residuals from the past 5 cycles as input and outputs the covariance against process noise. The correction amount is used to adaptively compensate for model mismatch caused by aircraft maneuvers. The fused high-dimensional flight situation is refreshed at a frequency of 10Hz, including the precise position, velocity vector, health status, mission intent, and surrounding obstacle distribution of each aircraft.

[0048] In the above method, step S3 implements dynamic airspace cellular management, specifically as follows: the initial airspace division uses a hexagonal grid, covering the entire low-altitude area of ​​the city (0-300 meters in altitude). The side length L of each hexagonal cell is determined by the formula... Confirmed, among which The maximum permitted flight speed in this area. The safe interval time (typically 5 seconds) is calculated as follows: Each cell is assigned a unique ID and associated with a set of access rules, including allowed altitude layers, maximum speed, flight direction (clockwise / counterclockwise loop), and priority level.

[0049] When the edge digital twin engine detects real-time traffic density within a cell When the preset density threshold is exceeded, a cell division mechanism is triggered. The division process is executed by the cell management service module: First, the original cell is divided into 7 smaller sub-cells, with the side length of the new sub-cells being 1 / 3 of the original side length; second, based on the current aircraft distribution heatmap, differentiated passage rules are assigned to each sub-cell, for example, the central sub-cell is set as a low-speed buffer, and the outer sub-cells are assigned unidirectional channels according to the direction of entry and exit; finally, the updated cell topology and rule set are packaged into a policy update message. This message is broadcast through the control plane channel of the 5G-A network slice and sent to all aircraft located in or about to enter the area via multicast. After receiving the policy update, the communication and sensing integrated module of the aircraft terminal immediately parses and loads the new airspace constraints, updating the feasible domain of the local trajectory planner. The cell merging mechanism is triggered after the traffic density has been below the threshold for 30 seconds to reduce management overhead. In addition, when meteorological data indicates that a certain area has experienced heavy rainfall ( ) or strong winds ( When the security level of a sensitive ground area is raised, the access rules of the corresponding cells above it are automatically tightened, for example, the maximum speed is reduced to 20 km / h.

[0050] In the above method, step S4 involves conflict prediction and trajectory pre-optimization. The specific process is as follows: The conflict prediction module within the edge digital twin engine uses the current high-dimensional flight situation as initial conditions and employs a lightweight graph neural network (GNN) to extrapolate traffic flow over the next 30 seconds. This GNN contains three graph convolutional layers, and the node feature vector is [position]. , , ,speed The feature vector of the edge [task priority p] is [relative distance d, proximity rate r], where the proximity rate The GNN outputs the probability distribution of each aircraft's position at each future time. The conflict resolution logic is as follows: if any two aircraft have a conflict at time within the prediction period... Make If it is, then it is determined to be a potential conflict. for Minimum prediction interval at time, The safe distance threshold is set according to the type of aircraft; 50 meters for drones and 100 meters for eVTOL. This is the separation velocity threshold (typically 2 m / s).

[0051] Once a potential conflict is detected, the trajectory optimization module initiates a four-dimensional collaborative decoupling calculation. This module constructs a joint optimization problem, targeting 1 to N affected aircraft. The optimization variable is the sequence of track points for each aircraft over the next T seconds. The constraints include: dynamic constraints (acceleration not exceeding...) ), spatial cell access rules, and minimum power maintenance requirements. Optimize the objective function. Defined as:

[0052]

[0053] in This is the additional energy consumption increment (relative to the original planned flight path). Delays caused by the task For the first The acceleration vector of the segment trajectory, Number of segments for the track, , , Normalized weighting coefficients ( The weighting coefficients are dynamically adjusted based on the task type: logistics and delivery tasks emphasize timeliness (…). =0.6), manned eVTOL focuses on comfort ( =0.7), emergency rescue missions focus on energy consumption ( =0.8). The optimization problem is solved using an improved particle swarm optimization (PSO) algorithm, which evaluates multiple candidate tracks in parallel in a digital twin environment, and finally outputs 1 to 3 four-dimensional cooperative decoupling tracks that satisfy all constraints. Each track contains a precise time-space coordinate sequence and a corresponding control command sequence.

[0054] In the above method, step S5, issuing structured guidance instructions, is specifically implemented as follows: the optimized four-dimensional track is converted into structured instruction packets, following the extended format of the ISO 21384-3 standard. Each instruction packet contains the following fields: timestamp T (UTC time, precision 1 millisecond), target velocity V (m / s), target altitude H (m), and three-dimensional waypoint coordinates (…). The instruction packet includes WGS-84 latitude, longitude, and altitude, instruction validity period (usually 5 seconds), and digital signature (for integrity verification). The instruction packet is sent through a dedicated control channel of the 5G-A network slice. This channel is configured with the highest priority QCI value of 1, uses SC-FDMA uplink and OFDMA downlink waveforms, and has a subcarrier spacing of 60kHz to reduce latency.

[0055] After receiving commands at the physical layer, the integrated communication and sensing module of the aircraft terminal quickly decapsulates them at the MAC layer and sends them to the application layer. The command parsing module completes syntax verification and semantic parsing within 2 milliseconds, generating low-level control commands that the flight control system can recognize. The flight control system's execution response time is less than 50 milliseconds, which is the time interval from the completion of command parsing to the start of action of the actuators (motors, control surfaces). For manned eVTOL, the system displays an operator confirmation interface before execution, providing a 3-second countdown; if the operator does not intervene, it executes automatically. All command execution statuses (success / failure / partial execution) are transmitted back to the edge computing node in real time via the uplink for closed-loop verification and digital twin status updates.

[0056] To verify the effectiveness of the above method, the following specific application example is constructed: At 10:00 AM in the central business district (CBD) of a first-tier city, a sudden fire occurs, requiring the urgent dispatch of a medical supply transport drone (model: DJI Motrice 350 RTK modified version) from a warehouse 5 kilometers away to the fire site. Simultaneously, 12 logistics drones, 3 eVTOL air taxis, and 2 power line inspection drones are operating in the same airspace.

[0057] The system execution flow is as follows: In step S1, the central control cloud platform receives an emergency instruction from the fire department, temporarily opens a green rescue channel from the warehouse to the fire scene, and raises the priority of all airspace cells within this channel to the highest level. The edge computing node loads the latest 3D building model (including smoke diffusion simulation data of the fire scene). In step S2, multi-level fusion perception detects thermal airflow disturbances above the fire scene, causing the VIBI index of an inspection drone to surge to 0.08, which the system marks as a "potential risk of loss of control". At the same time, side-link data shows that there is a temporary construction crane not marked in the GIS 50 meters ahead of a logistics drone. In step S3, the system divides the rescue channel into 5 dynamic cells, sets one-way traffic rules, and prohibits non-rescue aircraft from entering. In step S4, the conflict prediction module detects a risk of collision between the rescue drone and an eVTOL at an intersection after 30 seconds. Rice (meters), immediately initiate trajectory optimization. Objective function weights are set to... , =0.6, =0.1, generating two escape paths: Option A causes the eVTOL to decelerate and circle, Option B causes the rescue drone to veer slightly to the right. The system selects Option B to minimize overall delay. In step S5, the structured instruction " , "The command is sent to the rescue drone within 8 milliseconds, and the flight control system completes the execution within 45 milliseconds. The end-to-end latency from risk identification to command execution is 120 milliseconds, which is far lower than the more than 2 seconds required by the traditional alarm mode."

[0058] Example 2

[0059] In another application scenario, for security airspace control during large-scale sporting events, the system performs coordinated guidance of densely formation aircraft. During the event, the low-altitude airspace is divided into a core area (above the venue), a buffer zone (within a 2-kilometer radius), and an outer area. Only official aerial photography drones are allowed in the core area, the buffer zone restricts flight altitude to below 60 meters, and the outer area operates normally.

[0060] In step S1, the central control cloud platform preloads the event security plan, including dynamic no-fly zone expansion rules (such as automatically expanding the no-fly radius in case of spectator disturbances). Edge computing nodes deploy dedicated AI models to identify abnormal flight behavior (such as high-speed dives and low-altitude hovering). In step S2, multi-layered fusion sensing utilizes base stations... Passive positioning of unauthorized intruding consumer drones (even with their communication modules disabled) achieved a positioning accuracy of 25 centimeters. Simultaneously, CSI analysis revealed abnormal fluctuations in the propeller speed of a certain drone. The system determines that a hijacking is possible. In step S3, the system reduces the side length of the buffer cell from 35 meters to 20 meters, forming a finer-grained control grid. In step S4, the conflict prediction module detects that three official aerial photography drones are about to enter the same cell due to shooting angle requirements, posing a risk of close encounter. The trajectory optimization module prioritizes comfort. A smooth staggered entry scheme is generated, with each drone passing through key points at 1.5-second intervals. In step S5, the structured instructions include timestamps accurate to 0.1 seconds to ensure synchronized formation movements. Real-world testing shows that even with 200 drones operating concurrently, the system still maintains 99.999% instruction reliability and an average end-to-end latency of 8 milliseconds.

[0061] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for real-time monitoring and guidance of low-altitude aircraft based on 5G-A, characterized in that, Includes the following steps: Step S1: Construct a cloud-network-edge-device collaborative architecture, deploy global airspace management strategies and AI training models on the central control cloud platform, deploy edge computing nodes and run a low-altitude digital twin engine on the 5G-A ground base station side, integrate a communication and sensing integrated module on the aircraft terminal, and achieve highly reliable and low-latency data interaction between layers through low-altitude dedicated network slicing. Step S2: Perform multi-level fusion perception, use multiple 5G-A base stations to perform passive three-dimensional positioning of the target based on the signal arrival time difference and arrival angle, analyze the channel state information of the aircraft's uplink signal to monitor its flight status non-intrusively, and exchange local obstacle perception data through the side links between aircraft. The base station perception data, aircraft reported data and external environment data are spatiotemporally aligned and fused in the edge digital twin to generate a high-dimensional flight situation. Step S3: Implement dynamic airspace cell management, divide the low-altitude airspace into non-uniform virtual airspace cells, dynamically adjust the geometric boundaries and passage rules of each cell according to real-time traffic density, meteorological conditions and the status of sensitive ground areas, and broadcast the updated airspace strategy to relevant aircraft through 5G-A network slicing. Step S4: Perform conflict prediction and trajectory pre-optimization. Based on the real-time situation in the edge digital twin, use a lightweight AI model to extrapolate traffic flow in a specific time period in the future, identify potential conflicts or violation risks, and combine the mission type, performance constraints and remaining energy of the affected aircraft to generate one or more four-dimensional collaborative escape trajectories in the digital twin environment. Step S5: Issue structured guidance instructions, converting the optimized track into structured instructions containing time, speed, altitude and waypoint coordinates. These instructions are then sent to the relevant aircraft flight control system via the 5G-A control channel with end-to-end latency less than a preset threshold and reliability not lower than a preset reliability, ensuring quality of service. The flight control system can then directly parse and execute the instructions or execute them after operator confirmation.

2. The method for real-time monitoring and guidance of low-altitude aircraft based on 5G-A according to claim 1, characterized in that, The central control cloud platform connects bidirectionally with the civil aviation supervision system and the public security low-altitude security platform through a standardized API interface, receiving macro-level airspace opening instructions, no-fly zone updates, and emergency response commands, and simultaneously uploading anonymized flight activity statistics. The edge computing nodes are distributed and deployed on each 5G-A base station side, running a dynamic three-dimensional airspace mirror with horizontal accuracy of ±10 cm and vertical accuracy of ±5 cm. The integrated communication and sensing module supports the transmission and reception of modulated sensing signals for detecting surrounding obstacles.

3. The method for real-time monitoring and guidance of low-altitude aircraft based on 5G-A according to claim 1, characterized in that, The accuracy of the signal arrival time difference positioning is better than 30 centimeters, and the arrival angle resolution reaches 0.5 degrees. The channel state information analysis extracts the time delay spread and Doppler frequency shift characteristics of the multipath components to construct the aircraft vibration intensity index, which is the average value of the product of the time delay deviation and Doppler frequency shift of each effective multipath component. When the vibration intensity index exceeds a preset threshold, the aircraft is determined to be in an abnormal vibration state.

4. The method for real-time monitoring and guidance of low-altitude aircraft based on 5G-A according to claim 1, characterized in that, The inter-vehicle side walkway adopts a distributed scheduling mechanism based on a resource pool. The perception data exchange cycle is no more than 100 milliseconds. The exchanged content includes obstacle point cloud coordinates, relative distance, and confidence weight. The confidence weight is calculated based on the obstacle distance and relative azimuth angle, with obstacles in front receiving a higher confidence score.

5. The method for real-time monitoring and guidance of low-altitude aircraft based on 5G-A according to claim 1, characterized in that, The multi-source data fusion adopts a joint correction mechanism of Kalman filtering and deep learning. The state vector includes position, velocity, attitude angle and intention label. The observation noise covariance matrix is ​​dynamically adjusted according to the signal-to-noise ratio. The lower the signal-to-noise ratio, the larger the observation noise variance in the corresponding direction. The update step of Kalman filtering introduces a lightweight LSTM network, which takes the historical state residual as input and outputs the correction amount for the process noise covariance.

6. The method for real-time monitoring and guidance of low-altitude aircraft based on 5G-A according to claim 1, characterized in that, The initial division of the airspace cells is based on a hexagonal grid, with the side length determined by the maximum allowable flight speed and the safe interval time. When the local traffic density exceeds a preset density threshold, the cell automatically divides into multiple sub-cells. When weather conditions deteriorate or the security level of sensitive ground areas is upgraded, the passage rules of the relevant cells are automatically tightened.

7. The method for real-time monitoring and guidance of low-altitude aircraft based on 5G-A according to claim 1, characterized in that, The conflict prediction model adopts a graph neural network architecture. The node features include the aircraft position, velocity vector and mission priority, and the edge features are relative distance and proximity rate. The conflict determination condition is that the predicted minimum distance is less than the safe distance threshold and the separation speed is greater than the separation speed threshold.

8. The method for real-time monitoring and guidance of low-altitude aircraft based on 5G-A according to claim 1, characterized in that, The objective function for the four-dimensional trajectory optimization comprehensively considers the additional energy consumption increment, mission delay time, and the sum of squared accelerations; the weight coefficients are dynamically adjusted according to the mission type, with logistics and delivery missions emphasizing timeliness, manned aircraft emphasizing comfort, and emergency rescue missions emphasizing energy consumption.

9. The method for real-time monitoring and guidance of low-altitude aircraft based on 5G-A according to claim 1, characterized in that, The structured command format includes timestamps, target speed, target altitude, and three-dimensional waypoint coordinates, and follows the ISO21384-3 standard extension. Commands are issued through the 5G-A control channel with the highest priority QCI value, and the flight control system's execution response time is less than 50 milliseconds. For manned aircraft, an operator confirmation interface is provided before command execution, and the command is executed automatically if no intervention is required.

Images