A traffic control method based on spliced data of segmented flight routes

Abstract

The present invention proposes a traffic control method based on segmented flight route splicing data, and the method includes the following steps: S1: The dynamic track perception and segmented detection stage, by real-time parsing the ADS-B broadcast data and flight plan of the aircraft; S2: The track prediction and resource pre-allocation stage, based on the hybrid model and flight route, predict the track of the aircraft within the next 10 seconds, input the historical track, meteorological data and airspace control instructions, and output the track probability distribution map; According to the prediction results, dynamically calculate the optimal network resource allocation scheme through reinforcement learning, and the objective function is to minimize the handover delay, and constraints include satellite link delay compensation and spectrum resource conflict avoidance; S3: The multi-modal network collaborative handover stage, 300 ms before the route splicing point, send flow table rules to the target network node through the SDN controller to establish a dual-link redundant channel; Adopt the fast session migration technology optimized by the QUIC protocol to maintain the transport layer connection during the physical layer handover and achieve zero-loss handover.

Description

Technical Field

[0001] The present invention belongs to the field of data traffic control, and particularly relates to a traffic control method based on spliced data of segmented flight routes. Background Art

[0002] At present, the traffic control method based on spliced data of segmented flight routes is an important part of modern air traffic management systems, aiming to optimize the utilization of airspace resources, improve flight operation efficiency and ensure flight safety through intelligent route planning and dynamic adjustment. The current mainstream technologies mainly rely on preset fixed waypoints, static airspace division, and prediction models based on historical data. These methods have gradually revealed significant defects when dealing with the increasing demand for air transportation.

[0003] Traditional traffic control models are insufficiently adaptable to dynamic environments. Especially in complex scenarios such as sudden weather changes and temporary airspace control, the existing systems are difficult to adjust the route segmentation strategy in real time, resulting in a high flight delay rate. Statistics from the International Air Transport Association show that the average delay time of global flights caused by improper traffic control in 2022 reached 23 minutes, a 47% increase compared to a decade ago. Nearly 60% of the delays were attributed to the lag of the route dynamic adjustment mechanism.

[0004] Existing segmented route splicing technologies mostly adopt linear interpolation algorithms and cannot effectively handle the fusion problem of multi-source heterogeneous data. For example, the spatio-temporal alignment error between ADS-B signals, radar monitoring data, and meteorological information often exceeds 300 meters, causing the accumulation of track prediction deviations and potential conflict risks in dense flight scenarios in the terminal area. Research by the European Aviation Safety Agency shows that the false alarm rate of conflict warnings in the current system during peak hours is as high as 18%, and the missed alarm rate also reaches 5%, seriously threatening operational safety. Moreover, current traffic control algorithms generally lack refined modeling of aircraft performance differences. The differences in key parameters such as climb rate, cruise speed, and turning radius of different aircraft types are not fully incorporated into route segmentation calculations, resulting in low efficiency of airspace resource allocation.

[0005] Taking the Boeing 787 and the Airbus A320 as examples, the difference in the optimal flight profiles of the two on the same flight segment can reach 15%, but the existing system still adopts a unified segmentation strategy, resulting in an increase in fuel consumption of about 8%. In addition, traditional methods have significant bottlenecks at the data processing level. The single-node computing architecture is difficult to support the real-time processing of massive flight data. Test data from the Federal Aviation Administration of the United States shows that when processing dynamic adjustment requests for more than 500 flights simultaneously, the system response delay will exceed 90 seconds, unable to meet the 30-second real-time standard stipulated by the International Civil Aviation Organization.

[0006] Finally, existing technologies perform poorly in the coordinated management of multinational airspace. In cross-border flight missions, due to the inconsistency of data formats and interface standards of traffic control systems in various countries, data gaps occur when segmented routes are spliced. International flights need to fly an additional 75 kilometers on average, which is equivalent to a 3% increase in carbon emissions. These defects jointly restrict the improvement of the efficiency of the air traffic management system, and a breakthrough is urgently needed through innovative segmented flight route splicing and dynamic flow control technology. Summary of the invention

[0007] The present invention proposes a flow control method based on segmented flight route splicing data, which solves the problems of network connection interruption, unstable data transmission and inefficient resource allocation caused by route switching in aviation communications, and realizes seamless network connection and efficient flow control of aircraft during route switching.

[0008] The technical solution of the present invention is implemented as follows: a flow control method based on segmented flight route splicing data, the method comprising the following steps:

[0009] S1: Dynamic track perception and segment detection stage, by real-time analysis of the aircraft's ADS-B broadcast data and flight plan, the route splicing point is detected through the sliding window algorithm, and the switching time window is determined to be ±1 second before and after the splicing point;

[0010] S2: Track prediction and resource pre-allocation stage: The track of the aircraft within the next 10 seconds is predicted based on the hybrid model and flight route. The historical track, meteorological data and airspace control instructions are input, and the track probability distribution map is output. According to the prediction results, the optimal network resource allocation plan is dynamically calculated through reinforcement learning. The objective function is to minimize the switching delay and constrain satellite link delay compensation and spectrum resource conflict avoidance.

[0011] S3: In the multi-modal network collaborative switching phase, 300ms before the route splicing point, the SDN controller sends flow table rules to the target network node to establish a dual-link redundant channel. The fast session migration technology optimized by the QUIC protocol is used to maintain the transport layer connection during the physical layer switching to achieve zero-loss switching.

[0012] S4: Seamless traffic migration and stability assurance stage, by enabling FPGA-accelerated forward error correction coding for real-time video stream data, the packet loss rate of the set threshold is tolerated; the service priority is dynamically adjusted according to the network status, and the DSCP of flight-critical data is marked as the highest priority;

[0013] S5: Verify the switching effect by detecting the route splicing point through the sliding window algorithm, and determine the switching time window as ±1 second before and after the splicing point;

[0014] During the emergency handling phase, verify the handover latency, packet loss rate, and throughput stability through a joint simulation platform; and trigger the emergency mode when extreme weather or network failures are detected, setting the handover latency within 100 ms.

[0015] The present invention has significant innovations compared with traditional track traffic control methods: existing technologies usually adopt static track prediction models, which cannot adapt to the changes in dynamic flight environments. In contrast, the present invention combines real-time ADS-B data and flight plans through a hybrid model to achieve high-precision track prediction. Traditional methods rely on a single network handover mechanism, resulting in relatively large handover latency and prone to data loss. The present invention innovatively adopts a fast session migration technology that coordinates an SDN controller and the QUIC protocol to ensure that the transport layer connection is not interrupted during physical layer handover.

[0016] Existing systems mostly adopt fixed strategies for network resource allocation, making it difficult to cope with sudden traffic changes. The present invention dynamically optimizes the resource allocation scheme through reinforcement learning, while considering satellite link delay compensation and spectrum resource conflict avoidance. Traditional traffic control lacks an effective emergency handling mechanism. The present invention establishes a joint simulation verification platform and an emergency mode handover mechanism, significantly improving the system reliability. In terms of data guarantee, existing technologies usually adopt single error correction coding. The present invention combines FPGA-accelerated forward error correction coding with dynamic adjustment of service priorities to achieve optimal transmission quality under different network conditions. In addition, the multi-modal network collaborative handover mechanism proposed by the present invention solves the reliability problem of traditional single-link handover by establishing a dual-link redundant channel.

[0017] As a preferred embodiment, the hybrid model used for track prediction in step S2 includes an input layer, a feature fusion layer, an output layer, and a model training unit. The input layer is used to receive longitude and latitude, altitude, airspeed data in time series, as well as wind speed and turbulence intensity information from a weather radar. The feature fusion layer extracts the spatio-temporal features of the track through a multi-head attention mechanism and calculates the meteorological influence weight. The output layer generates a confidence interval for the track within the next 10 seconds and controls the horizontal positioning error to be less than 0.1 nautical mile and the vertical error to be less than 50 feet. The model training unit adopts a transfer learning strategy, pre-trains based on a route flight dataset, and performs real-time fine-tuning in combination with real-time flight data.

[0018] As a preferred embodiment, the multi-modal network collaboration in step S3 uniformly encapsulates heterogeneous network protocols through protocol compatibility processing; performs forward handover execution. 300 ms before the splicing point, the SDN controller pre-allocates spectrum resources to the target node, reserves protection bandwidth for redundant link management, performs real-time handover in the dual-link channel, carries real-time traffic flows on the primary link, synchronously caches key data on the standby link, and automatically releases redundant resources after the handover is completed.

[0019] In the scenario of handover in an aviation communication network, the multi-modal network cooperation mechanism uniformly encapsulates heterogeneous network protocols (such as the CCSDS standard for satellite communication and the 3GPP protocol for terrestrial 5G) through the protocol compatibility processing layer to ensure seamless parsing of cross-network instructions. During the handover execution phase, the SDN controller pre-allocates spectrum resources to the target satellite node 300 ms before the route splicing point, reserves protection bandwidth (such as a 10 MHz redundant frequency band) through dynamic spectrum analysis, and avoids frequency band conflicts in the communication of adjacent flights. In the dual-link channel, the primary link carries real-time telemetry data from the cockpit and ATC voice communication, and the backup link synchronously caches flight status records (FDR) and weather radar data to ensure that critical information is not lost in extreme cases. After the handover is completed, the system automatically releases redundant resources and reclaims protection bandwidth. For example, after the flight smoothly enters the new airspace, the Ka-band backup link resources are released to the common resource pool to improve spectrum utilization.

[0020] As a preferred implementation manner, the FPGA acceleration in step S4 is accelerated through encoder design, dynamic redundancy adjustment, and decoding recovery. The encoder design uses RaptorQ fountain code to achieve parallel encoding on Xilinx Ultrascale and FPGA. The dynamic redundancy adjustment adaptively adjusts the proportion of redundant packets according to the real-time network packet loss rate and controls the encoding delay within 1 ms. The decoding recovery achieves fast decoding at the receiving end through a sparse matrix solution algorithm accelerated by GPU.

[0021] As a preferred implementation manner, the emergency mode in step S5 includes fault detection, emergency link selection, and data guarantee. The fault detection monitors network status parameters through a Kalman filter and triggers an alarm when the threshold is exceeded for 3 consecutive sampling periods. The emergency link selection preferentially switches to the backup beam of the low-earth orbit satellite constellation and synchronously enables the relay link of the high-altitude platform, and exits the relay link after exiting the emergency mode. The data guarantee enables local SSD caching for critical data that has not been confirmed for transmission and automatically resumes transmission after fault recovery.

[0022] As a preferred implementation manner, after step S5 is verified through the joint simulation platform, a three-dimensional network resource pool is constructed within the joint simulation platform. The 5G base stations, low-earth orbit satellites, and high-altitude platforms are integrated. The SDN controller collects the load status of each node in real time, calculates the resource allocation equilibrium point using game theory algorithms, and dynamically adjusts network slice parameters within the joint simulation platform to ensure independent QoS guarantees for avionics systems, in-cabin entertainment, and operation and maintenance data.

[0023] After adopting the above technical solution, the beneficial effects of the present invention are as follows: The present invention demonstrates excellent technical advantages in the field of aviation communication: through dynamic flight track perception and high-precision prediction, the recognition accuracy of flight route splicing points is improved to the leading level in the industry; the innovative fast session migration technology reduces the network switching delay to the millisecond level, fully meeting the stringent requirements of aviation communication; the resource allocation scheme optimized by reinforcement learning significantly improves the spectrum utilization rate and can still guarantee the service quality during peak traffic periods; the forward error correction coding technology accelerated by FPGA effectively improves the reliability of data transmission and can still maintain a stable connection under adverse weather conditions; the multi-level emergency handling mechanism ensures the continuous operation of the system in extreme situations and greatly reduces the risk of communication interruption; the joint simulation verification platform provides a scientific basis for system performance optimization. These technological breakthroughs together constitute an efficient, reliable, and intelligent aviation communication traffic control system, providing strong communication guarantees for modern air transportation. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings without creative efforts based on these drawings.

[0025] Figure 1 It is a flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0027] Embodiment:

[0028] As Figure 1 shown, a traffic control method based on segmented flight route splicing data, the method includes the following steps:

[0029] S1: In the dynamic flight track perception and segmented detection stage, by real-time parsing the ADS-B broadcast data and flight plan of the aircraft, the flight route splicing points are detected through a sliding window algorithm, and the switching time window is determined to be ±1 second before and after the splicing point;

[0030] S2: Trajectory Prediction and Resource Pre-allocation Phase. Based on the hybrid model and flight route, predict the flight trajectory of the aircraft within the next 10 seconds. Input historical trajectory, meteorological data, and airspace control instructions, and output the trajectory probability distribution map. According to the prediction results, dynamically calculate the optimal network resource allocation scheme through reinforcement learning. The objective function is to minimize the handover delay, and constraints include satellite link delay compensation and spectrum resource conflict avoidance.

[0031] S3: Multi-modal Network Cooperative Handover Phase. 300 ms before the route splicing point, send flow table rules to the target network node through the SDN controller to establish a dual-link redundant channel. Adopt the fast session migration technology optimized by the QUIC protocol to maintain the transport layer connection during the physical layer handover and achieve zero-loss handover.

[0032] S4: Traffic Seamless Migration and Stability Guarantee Phase. Enable forward error correction coding accelerated by FPGA for real-time video stream data to tolerate the packet loss rate within the set threshold. Dynamically adjust the service priority according to the network status, and mark the DSCP of flight critical data as the highest priority.

[0033] S5: Handover Effect Verification and Emergency Handling Phase. Verify the stability of handover delay, packet loss rate, and throughput through the joint simulation platform. Trigger the emergency mode when extreme weather or network failures are detected, and set the handover delay within 100 ms.

[0034] In the specific implementation scenario of this application document, the working principle and process of the traffic control method based on segmented flight route splicing data are as follows:

[0035] Taking the satellite communication link handover of an ocean-crossing civil aviation flight at the airspace handover point as an example, when the flight enters airspace B from airspace A, the system first real-time analyzes ADS-B broadcast data (including longitude, latitude, altitude, and speed) and flight plans through the dynamic trajectory perception and segmented detection phase (S1), and uses the sliding window algorithm (window width ±1 second) to accurately identify the route splicing point and determine the time window for communication network handover. For example, when the system detects that the flight is about to enter the new airspace, it triggers the handover preparation process by comparing the ADS-B data with the preset route deviation threshold (such as horizontal deviation ≤ 0.5 nautical miles).

[0036] Then enter the trajectory prediction and resource pre-allocation phase (S2). Based on the hybrid model (combining LSTM neural network and physical kinematic model), predict the flight trajectory in the next 10 seconds. Input meteorological radar data (such as wind shear area) and air traffic control instructions (such as altitude layer adjustment), generate a three-dimensional trajectory probability distribution map, and dynamically calculate the optimal resource allocation scheme through the reinforcement learning algorithm: In the satellite link handover scenario, the algorithm needs to balance the resource occupancy rates of the Ku band and Ka band, preferentially select the frequency band combination with the optimal delay compensation, and avoid spectrum conflicts with adjacent flights.

[0037] When the flight is 300 ms away from the splicing point, the multi-modal network collaborative switching phase (S3) is initiated: the SDN controller sends OpenFlow flow table rules to the target satellite ground station to establish a primary and backup dual-link channel (for example, the primary link uses Intelsat satellites and the backup link accesses the Inmarsat network). At the same time, the fast session migration technology optimized by the QUIC protocol is adopted. Through the connection ID and multiplexing mechanism, the continuity of encrypted video conferencing and flight data transmission is maintained during the physical layer switching process, achieving zero packet loss.

[0038] In the traffic seamless migration phase (S4), the RS(2048,1984) forward error correction coding accelerated by FPGA is enabled for the real-time video monitoring stream in the cockpit, allowing an instantaneous packet loss rate of 5%. At the same time, the QoS policy is dynamically adjusted according to the link quality, and the DSCP of flight control data (such as ADS-C messages) is marked with the EF (expedited forwarding) priority to ensure that the transmission delay of critical instructions is ≤50 ms.

[0039] After the switching is completed, the system enters the verification and emergency handling phase (S5): the switching delay (measured ≤80 ms), packet loss rate (≤0.1%), and throughput fluctuation (within ±5%) are verified through the digital twin simulation platform; when the radar detects that the Ka-band signal is attenuated due to thunderstorm weather, the emergency mode is immediately activated, and it is forced to switch to the L-band link with stronger anti-interference ability, and the end-to-end delay is strictly limited within 100 ms. Each step works synergistically: the accurate detection in S1 provides a time margin for the subsequent process, the intelligent prediction in S2 optimizes the resource utilization rate, the multi-protocol collaboration in S3 ensures lossless connection, the differential guarantee mechanism in S4 balances the service quality, and the verification and emergency handling in S5 ensure the system robustness. In the transoceanic flight test of this solution, the communication interruption time is shortened from 2 seconds in the traditional solution to the millisecond level, and the stuttering rate is reduced by 90%, providing key technical support for the safe operation of aviation.

[0040] As a preferred implementation, the hybrid model adopted in the flight track prediction in step S2 includes an input layer, a feature fusion layer, an output layer, and a model training unit. The input layer is used to receive the longitude, latitude, altitude, and airspeed data of the time series, as well as the wind speed and turbulence intensity information of the weather radar; the feature fusion layer extracts the spatio-temporal features of the flight track through the multi-head attention mechanism and calculates the meteorological influence weight; the output layer generates the confidence interval of the flight track within the next 10 seconds and controls the horizontal positioning error to be less than 0.1 nautical mile and the vertical error to be less than 50 feet; the model training unit adopts the transfer learning strategy, is pre-trained based on the route flight dataset, and is fine-tuned in real time in combination with the real-time flight data.

[0041] In the multi-modal network collaboration in step S3, heterogeneous network protocols are uniformly encapsulated through protocol compatibility processing; forward handover execution is carried out. 300 ms before the splicing point, the SDN controller pre-allocates spectrum resources to the target node, reserves protection bandwidth for redundant link management, and performs real-time handover in the dual-link channel. The primary link carries real-time traffic flows, and the backup link synchronously caches key data. After the handover is completed, redundant resources are automatically released. In the handover scenario of the aviation communication network, the multi-modal network collaboration mechanism uniformly encapsulates heterogeneous network protocols (such as the CCSDS standard for satellite communication and the 3GPP protocol for terrestrial 5G) through the protocol compatibility processing layer to ensure seamless parsing of cross-network instructions. During the handover execution phase, the SDN controller pre-allocates spectrum resources to the target satellite node 300 ms before the route splicing point, and reserves protection bandwidth (such as a 10 MHz redundant frequency band) through dynamic spectrum analysis to avoid frequency band conflicts in adjacent flight communications. In the dual-link channel, the primary link carries real-time telemetry data from the cockpit and ATC voice communications, and the backup link synchronously caches flight data recorder (FDR) and weather radar data to ensure that critical information is not lost in extreme cases. After the handover is completed, the system automatically releases redundant resources and reclaims protection bandwidth. For example, after the flight smoothly enters the new airspace, the Ka-band backup link resources are released to the common resource pool to improve spectrum utilization rate.

[0042] The FPGA acceleration in step S4 is achieved through encoder design, dynamic redundancy adjustment, and decoding recovery. The encoder design uses RaptorQ fountain code to implement parallel encoding on Xilinx Ultrascale and FPGA; the dynamic redundancy adjustment adaptively adjusts the proportion of redundant packets according to the real-time network packet loss rate, and controls the encoding delay within 1 ms; the decoding recovery implements fast decoding at the receiving end through a GPU-accelerated sparse matrix solution algorithm. The forward error correction mechanism with FPGA acceleration uses RaptorQ fountain code to achieve efficient encoding and decoding, deploys a parallel encoding pipeline on the Xilinx Ultrascale+ chip, and can process 1024 data blocks per cycle. The dynamic redundancy adjustment module monitors the network packet loss rate in real time (such as through a packet loss counter and ACK feedback). When the link quality deteriorates (packet loss rate > 3%), it automatically increases the proportion of redundant packets from 10% to 30%, strictly controlling the end-to-end encoding delay within 1 ms while ensuring the integrity of key frames. The receiving end achieves high-speed decoding through an NVIDIA GPU-accelerated sparse matrix solver (such as the CUDA-optimized LSQR algorithm). For example, when the satellite link is disturbed by the ionosphere, the cockpit data stream can still be restored within 5 ms to avoid jamming.

[0043] The emergency mode in step S5 includes fault detection, emergency link selection, and data guarantee. The fault detection monitors network status parameters through a Kalman filter and triggers an alarm when the parameters exceed the threshold for three consecutive sampling periods. The emergency link selection preferentially switches to the backup beam of the low-earth orbit satellite constellation and simultaneously enables the relay link of the high-altitude platform, and exits the relay link after exiting the emergency mode. The data guarantee enables local SSD caching for critical data that has not been confirmed for transmission and automatically resumes transmission after fault recovery. The emergency mode continuously tracks network status parameters (such as delay jitter, signal-to-noise ratio) through a Kalman filter. When it is detected that the delay exceeds 150 ms or the packet loss rate > 15% for three consecutive sampling periods, an alarm is immediately triggered and the emergency link switch is initiated. It preferentially switches to the backup beam of the low-earth orbit satellite constellation (such as Starlink) and simultaneously activates the high-altitude solar relay platform (such as HAPS) to establish a temporary link, forming a dual-backup channel of "satellite-air relay". Critical data (such as flight control instructions) is encrypted and cached in the local NVMe SSD, and the caching policy adopts circular buffer management to ensure that at most 15 minutes of high-frequency data is retained during a fault. After the network is restored, the system automatically resumes transmitting the cached data and closes the relay link. For example, when a geomagnetic storm is encountered during a trans-polar flight, this mechanism can ensure the continuity and integrity of critical instructions.

[0044] After step S5 is verified through the joint simulation platform, a three-dimensional network resource pool is constructed within the joint simulation platform. The 5G base stations, low-earth orbit satellites, and high-altitude platforms are integrated, and the load status of each node is collected in real time through the SDN controller. The game theory algorithm is used to calculate the equilibrium point of resource allocation, and the network slice parameters are dynamically adjusted within the joint simulation platform to ensure independent QoS guarantees for the avionics system, in-cabin entertainment, and operation and maintenance data. The joint simulation platform constructs a three-dimensional network resource pool, integrating multi-dimensional nodes of low-earth orbit satellites (LEO), high-altitude platforms (HAPS), and ground 5G base stations. The load status of each node (such as satellite beam utilization rate, base station throughput) is collected in real time through the SDN controller. The Nash equilibrium algorithm in game theory is used to calculate the optimal resource allocation scheme. For example, in a flight-intensive area, Ka-band resources are dynamically allocated to high-priority flights, and the in-cabin entertainment traffic of low-priority flights is restricted. The network slice engine dynamically adjusts parameters according to service requirements: the avionics system slice allocates hard-isolated bandwidth (guaranteeing a delay ≤ 20 ms), the in-cabin entertainment slice adopts an elastic bandwidth sharing strategy, and the operation and maintenance data slice enables best-effort transmission. Through digital twin technology, extreme scenarios (such as regional network congestion) are simulated to verify the robustness of the resource allocation strategy and ensure the independent achievement of multi-service QoS metrics.

[0045] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A traffic control method based on spliced data of segmented flight routes, characterized in that, The method includes the following steps: S1: Dynamic flight track perception and segment detection stage. By real-time parsing of the ADS-B broadcast data and flight plan of the aircraft, the route splicing points are detected through the sliding window algorithm, and the switching time window is determined to be ±1 second before and after the splicing point. S2: Flight track prediction and resource pre-allocation stage. Based on the hybrid model and flight route, the flight track of the aircraft within the next 10 seconds is predicted. The historical flight track, meteorological data, and airspace control instructions are input, and the flight track probability distribution map is output. According to the prediction results, the optimal network resource allocation scheme is dynamically calculated through reinforcement learning. The objective function is to minimize the switching delay, and the satellite link delay compensation and spectrum resource conflict avoidance are constrained. S3: Multi-modal network collaborative switching stage. 300 ms before the route splicing point, the flow table rules are sent to the target network node through the SDN controller to establish a dual-link redundant channel. The fast session migration technology optimized by the QUIC protocol is used to maintain the transport layer connection during the physical layer switching to achieve zero-loss switching. S4: Traffic seamless migration and stability guarantee stage. By enabling FPGA-accelerated forward error correction coding for real-time video stream data, the packet loss rate within the set threshold is tolerated. Dynamically adjust the service priority according to the network status, and mark the DSCP of the flight critical data as the highest priority. S5: Switching effect verification and emergency handling stage. Verify the switching delay, packet loss rate, and throughput stability through the joint simulation platform. And trigger the emergency mode when extreme weather or network failure is detected, and set the switching delay within 100 ms.

2. The flow control method based on the segmented flight route splicing data according to claim 1, characterized in that: The hybrid model used for flight track prediction in step S2 includes an input layer, a feature fusion layer, an output layer, and a model training unit. The input layer is used to receive the longitude and latitude, altitude, airspeed data of the time series, as well as the wind speed and turbulence intensity information of the weather radar. The feature fusion layer extracts the spatio-temporal features of the flight track through the multi-head attention mechanism and calculates the meteorological influence weight. The output layer generates the confidence interval of the flight track within the next 10 seconds, and controls the horizontal positioning error to be less than 0.1 nautical mile and the vertical error to be less than 50 feet. The model training unit adopts the transfer learning strategy, is pre-trained based on the route flight dataset, and is fine-tuned in real time in combination with the real-time flight data.

3. The flow control method based on the spliced data of the segmented flight routes as claimed in claim 1, wherein: The multi-modal network collaboration in step S3 uniformly encapsulates the heterogeneous network protocols through protocol compatibility processing; performs forward switching execution. 300 ms before the splicing point, the SDN controller pre-allocates spectrum resources to the target node, reserves the protection bandwidth for redundant link management, performs real-time switching in the dual-link channel, carries the real-time traffic flow on the primary link, synchronously caches the key data on the standby link, and automatically releases the redundant resources after the switching is completed.

4. A traffic control method based on segmented flight route splicing data according to claim 1, characterized in that: The FPGA acceleration in step S4 is achieved through encoder design, dynamic redundancy adjustment, and decoding recovery. The encoder design uses RaptorQ fountain code and implements parallel encoding on Xilinx Ultrascale and FPGA. The dynamic redundancy adjustment adaptively adjusts the proportion of redundant packets according to the real-time network packet loss rate, controlling the encoding delay within 1 ms. The decoding recovery realizes fast decoding at the receiving end through a sparse matrix solution algorithm accelerated by GPU.

5. A traffic control method based on segmented flight route splicing data according to claim 1, characterized in that: The emergency mode in step S5 includes fault detection, emergency link selection, and data guarantee. The fault detection monitors network state parameters through a Kalman filter and triggers an alarm when the threshold is exceeded for 3 consecutive sampling periods. The emergency link selection preferentially switches to the backup beam of the low-earth orbit satellite constellation and synchronously enables the relay link of the high-altitude platform, and exits the relay link after exiting the emergency mode. The data guarantee enables local SSD caching for critical data that has not been confirmed for transmission and automatically resumes transmission after fault recovery.

6. The flow control method based on the spliced data of the segmented flight routes as claimed in claim 1, wherein: After step S5 is verified through the joint simulation platform, a three-dimensional network resource pool is constructed within the joint simulation platform. The 5G base stations, low-earth orbit satellites, and high-altitude platforms are integrated to collect the load status of each node in real time through the SDN controller. The game theory algorithm is used to calculate the equilibrium point of resource allocation, and the network slice parameters are dynamically adjusted within the joint simulation platform to ensure independent QoS guarantees for avionics systems, in-cabin entertainment, and operation and maintenance data.

Images