Control tower and flight guarantee command room decision instruction transmission method based on edge nodes

Abstract

The invention belongs to the technical field of air traffic management, and discloses a control tower and flight guarantee command room decision instruction transmission method based on edge nodes, priorities are distributed for instructions through an intelligent grading model, the high-priority instructions can reserve resources in advance and are transmitted preferentially, the resources can be preempted in emergency, and the control tower and flight guarantee command room decision instruction transmission efficiency is improved. A multi-path transmission strategy is automatically switched when a certain path fails, data integrity is ensured in combination with fragment verification and block chain recording, instruction loss caused by a single point of failure is avoided through edge-cloud redundant backup, and the problems that instruction transmission is prone to interference and insufficient in safety in a traditional method are solved; through intelligent priority grading, resources are inclined to high-priority instructions, resource waste is avoided, a collaborative scheduling mechanism of an edge node cluster can dynamically adjust resource allocation according to a real-time load, and a transmission quality feedback and adaptive optimization strategy is continuously optimized, so that the system can still efficiently process a large number of instructions in a flight take-off and landing intensive period.

Description

Technical Field

[0001] This invention belongs to the field of air traffic management technology, specifically a method for transmitting decision commands in control towers and flight support command rooms based on edge nodes. Background Technology

[0002] In the field of air traffic control, the transmission of decision-making instructions between the control tower and the flight support command center must meet the requirements of high real-time performance, high reliability, and low latency to ensure the safety of flight takeoffs and landings and operational efficiency. Taking the development of digital collaborative air traffic control services in China's civil aviation as an example, although some progress has been made, such as the significant reduction in the time for pilots to obtain takeoff and landing information through digital information services in airport control towers, there are still shortcomings in the key links of decision-making instruction transmission.

[0003] In existing technologies, edge node clusters, priority classification, multipath transmission, and reinforcement learning are all independent technical means.

[0004] A search revealed that invention patent CN119342424A discloses a broadcast multicast (MBS) transmission method, device, terminal, and apparatus. This patent enables the tower to receive broadcast data packets from core network equipment and accurately transmit them to the terminal. Its advantage lies in completing broadcast multicast transmission without modifying the terminal, core network, or access network equipment. However, it only implements broadcast transmission and does not involve priority and resource linkage. In aviation scenarios, it is difficult to ensure the priority and rapid transmission of emergency instructions. Furthermore, relying on core network equipment, it cannot effectively offload and process traffic through local edge nodes when the network is congested. Relying on core network transmission without distinguishing instruction priorities makes it difficult to guarantee the priority transmission of emergency instructions, and it cannot meet the stringent requirements of aviation command for low latency and high reliability. Summary of the Invention

[0005] The purpose of this invention is to provide a method for transmitting decision-making instructions in control towers and flight support command rooms based on edge nodes, so as to solve the problems mentioned in the background art.

[0006] This invention organically integrates edge node cluster construction, priority classification, resource reservation, multi-path transmission (S4), redundancy backup, and adaptive optimization: priority classification results directly drive resource reservation and path selection, and edge node collaborative verification and redundancy backup are dynamically adjusted based on real-time link status. This end-to-end linkage mechanism is not available in existing single-module optimization technologies, and is particularly suitable for the high real-time and high reliability requirements of command transmission in aviation command scenarios. Compared with existing technologies, the prior art patent publication number CN119342424A focuses on ordinary broadcast multicast services and does not involve the relationship between priority classification and resource scheduling. In contrast, this invention, through priority-driven resource reservation and multi-path selection, dynamically matches command transmission efficiency with command importance, which is unique to aviation command scenarios.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for transmitting decision-making instructions in the control tower and flight support command center based on edge nodes, the specific steps of which are as follows:

[0008] S1. Edge Node Cluster Construction and Dynamic Networking: Deploy edge nodes containing NVIDIA Jetson chips, millimeter-wave communication modules and fiber optic interfaces in key areas of the airport. Build a distributed cluster through a self-organizing algorithm and use a sliding window algorithm to monitor link bandwidth, latency and packet loss rate in real time. When the indicators exceed the threshold, adjust the network topology using fiber optic backup links to establish a stable basic network for command transmission.

[0009] S2. Intelligent Priority Classification of Decision Instructions: Based on the established edge cluster, an MLP neural network model is run. Input features such as instruction type, flight status and time sensitivity are input. After nonlinear processing by the hidden layer, the output layer outputs four priority levels: emergency, high, medium and low through the softmax function.

[0010] S3. Priority-based edge node resource reservation: Based on the hierarchical results, the resource management module reserves CPU cores, contiguous memory space, and dedicated bandwidth tokens for high-priority instructions; in the event of an emergency instruction, a preemption mechanism is triggered to temporarily allocate low-priority resources so that critical instructions can obtain sufficient processing and transmission resources.

[0011] S4. Multi-path parallel transmission and dynamic route switching: After resource reservation is completed, medium and high priority instructions enable multi-path transmission. Paths are selected by hop count and link quality assessment. A monitoring thread tracks the status. When a path fails, a reselection algorithm switches to a high-quality path, so that instructions can flow efficiently in the cluster.

[0012] S5. Command content fragmentation and edge node collaborative verification: During transmission, large-capacity commands are fragmented according to their logical structure, and identifiers are generated using a hash algorithm and distributed to edge nodes. Nodes verify the integrity using a CRC algorithm, and the transmission information is recorded using a blockchain. After all fragments are received, the target node reassembles and performs a second verification to ensure the integrity of the command.

[0013] S6, Edge-Cloud Collaborative Command Redundancy Backup: After sharding verification, based on node load and network distance, command copies are synchronized to 2-3 nearby edge nodes, and erasure coding technology is used to prevent data loss; critical commands are synchronously uploaded to the cloud and long-term backup is achieved through object storage, establishing a two-layer security defense.

[0014] S7. Real-time feedback and adaptive optimization of transmission quality: Throughout the process, the target node calculates indicators such as latency and packet loss rate in real time and transmits them to the cluster via the feedback channel. The cluster initiates reinforcement learning based on the data, uses transmission performance as a reward signal, and dynamically adjusts the priority model, path strategy and resource allocation to continuously improve transmission quality.

[0015] Preferably, the specific steps for edge node cluster construction and dynamic networking in step S1 are as follows:

[0016] S11. Node Deployment and Hardware Configuration: Edge nodes are deployed in the control tower, flight support command center and key areas of the airport. The hardware uses NVIDIA Jetson series computing chips, whose powerful parallel computing capabilities can quickly process instruction data. Millimeter-wave communication modules are used to achieve high-speed wireless connection, and fiber optic interfaces are also provided as redundant backups. This configuration provides nodes with efficient processing and diverse communication capabilities, providing a hardware foundation for cluster construction.

[0017] S12. Self-organizing networking and link optimization: Based on the self-organizing algorithm, nodes establish initial connections through millimeter waves to form a distributed cluster. The sliding window algorithm is used to monitor link quality in real time. When indicators such as bandwidth and latency exceed the standard, dynamic topology adjustment is initiated, and the optical fiber link is used to replan the path. This mechanism keeps the cluster network stable.

[0018] Preferably, the specific steps of intelligent priority classification of decision instructions in step S2 are as follows:

[0019] S21. Evaluation Model Construction and Feature Input: Based on the distributed computing power of edge clusters, an MLP neural network priority evaluation model is built. The input layer integrates features such as instruction type, flight status and time sensitivity. For example, emergency avoidance instructions and flight correlation information during takeoff are included in the analysis to provide multi-dimensional data support for the model.

[0020] S22. Feature Processing and Priority Output: The model's hidden layers extract feature associations using non-linear activation functions, and the output layer uses a softmax function to generate a four-level priority probability distribution, automatically classifying instructions. This result accurately distinguishes the urgency of instructions, providing clear guidance for subsequent differentiated resource allocation and transmission strategies.

[0021] Formula for calculating softmax priority probability:

[0022]

[0023] In the formula: P(y=k|z) is the probability that the instruction belongs to the k-th priority level; z k is the output value of the k-th neuron in the output layer of the neural network; C is the total number of priority levels (C=4, corresponding to urgent, high, medium, and low); e is a natural constant;

[0024] The formula originates from the classic activation function of the output layer of a neural network. In this method, the input z is combined with aviation command features. k We performed weighted optimization to make the probability distribution more aligned with the priority determination requirements of aviation scenarios.

[0025] Preferably, the specific steps for priority-based edge node resource reservation in step S3 are as follows:

[0026] S31. Conventional resource reservation mechanism: Based on the priority results, the resource management module starts the reservation mechanism. The CPU uses the process scheduling algorithm to allocate independent cores for high-priority instructions. The memory reserves continuous space through paging technology. The bandwidth is allocated with dedicated tokens through the token bucket algorithm, thus configuring resources in advance for instructions of different priorities.

[0027] S32. Emergency Resource Scheduling Strategy: When an emergency command is detected, a preemption mechanism is triggered. Based on the priority strategy, some resources of low-priority commands are temporarily allocated to ensure that emergency commands can quickly obtain sufficient resources and avoid delays caused by insufficient resources.

[0028] Preferably, the specific steps of multi-path parallel transmission and dynamic route switching in step S4 are as follows:

[0029] S41. Multi-path detection and selection: After resource reservation is completed, multi-path transmission is initiated for medium and high priority commands. Through the link quality comprehensive scoring model, probe packets are sent periodically to collect latency and bandwidth information to build a status table, thereby filtering multiple available paths. The multi-path design provides multiple channels for command transmission and improves the flexibility of transmission.

[0030] The expression for the comprehensive link quality scoring model is as follows:

[0031]

[0032] Where: S is the overall path score (the higher the value, the better the path); α and β are weighting coefficients (α+β=1); H is the number of hops in the path; Q is the link quality parameter, a normalized value of the overall bandwidth, latency, and packet loss rate, in the range [0,1].

[0033] Formula source: Based on the improvement of the multi-index weighted model in the routing algorithm, and referring to the link evaluation method in "Design of Routing Protocols for AdHoc Networks", the dynamic adjustment mechanism of low latency weight β is strengthened for aviation scenarios;

[0034] S42. Real-time monitoring and dynamic switching: A dedicated thread is set up to monitor the path status. When a path experiences increased latency or failure, the traffic is switched to a high-quality path through a reselection algorithm, achieving seamless connection of command transmission. This enables medium and high priority commands to flow efficiently in the cluster and reduces the risk of transmission interruption.

[0035] Preferably, the specific steps for instruction content fragmentation and edge node collaborative verification in step S5 are as follows:

[0036] S51, Large-capacity instruction fragmentation: In multi-path transmission, large-capacity instructions are fragmented according to logical structure, split according to functional modules, and a unique identifier is generated by hash algorithm and distributed to different edge nodes. Fragmentation reduces the pressure of single-path transmission, and the identifier is used to realize the orderly management of data.

[0037] S52. Fragmentation Verification and Reassembly Verification: After receiving the fragments, each node uses the CRC algorithm to verify their integrity. Blockchain technology records the transmitted information to ensure that it cannot be tampered with. After all fragments are received, the target node reassembles them according to their identifiers and performs secondary verification, thereby comprehensively ensuring the integrity of the instruction data and avoiding information loss or tampering during transmission.

[0038] Preferably, the specific steps of the edge-cloud collaborative instruction redundancy backup in step S6 are as follows:

[0039] S61. Edge node local redundant storage: After completing the fragmentation verification, the instruction copy is stored on 2-3 neighboring nodes based on the DHT algorithm. The redundant fragments are encoded using erasure coding technology. By leveraging node collaboration, the instruction can be quickly recovered, which can cope with the risk of single point of failure and ensure the short-term security of data.

[0040] S62, Cloud-based Long-Term Backup Collaboration: For critical instructions, copies are synchronously uploaded to the cloud and object storage technology is used to achieve long-term preservation. Edge local backup and cloud backup work together to meet the needs of rapid recovery and achieve long-term data archiving, realizing multi-layered instruction security protection.

[0041] Preferably, the specific steps of real-time transmission quality feedback and adaptive optimization in step S7 are as follows:

[0042] S71, Full-process performance monitoring and feedback: Throughout the entire command transmission process, the target node's performance calculation module measures indicators such as latency and packet loss rate in real time, and feeds them back to the source node and cluster through a dedicated channel, providing the system with accurate transmission quality data;

[0043] S72. Reinforcement Learning Dynamic Optimization: Based on feedback data, the cluster uses reinforcement learning to build an optimization model. With transmission performance as the reward function, it adjusts variables such as priority models and path strategies. Through continuous iteration, it achieves dynamic optimization of strategies at each stage and continuously improves the overall transmission quality.

[0044] Reinforcement learning reward function formula:

[0045]

[0046] In the formula: R is the transmission quality reward value; k1, k2, and k3 are weighting coefficients representing the actual transmission delay; Dmax is the maximum allowable delay threshold; L is the packet loss rate; 1 is the integrity index, where 1 = 1 for complete transmission and 1 < 1 otherwise.

[0047] Formula source: Based on the reward mechanism design in reinforcement learning, and integrating the transmission quality evaluation index in "Adaptive Communication Network Optimization", the formula adapts to the optimization requirements of different instruction types by dynamically adjusting the k1-k3 weights.

[0048] The beneficial effects of this invention are as follows:

[0049] 1. This invention achieves local processing and transmission of decision-making instructions by deploying edge node clusters in key areas of the airport, eliminating the need for cloud relay. Simultaneously, the dynamic networking mechanism optimizes node connections in real time, and multi-path parallel transmission and dynamic routing switching avoid congested links, significantly shortening instruction transit time. This edge-side collaborative processing and transmission strategy effectively solves the distance bottleneck of traditional cloud transmission, significantly improving the real-time performance of instruction transmission and meeting the rapid response requirements for time-sensitive instructions in aviation command. The linkage between steps S2 (priority classification) and S3 (resource reservation) dynamically adjusts the allocation of resources based on instruction importance. The combination of S4 (multipath transmission) and S7 (reinforcement learning) allows for path strategy adjustment based on real-time link quality, forming a fully adaptive transmission mechanism. This collaborative design is an innovation not covered by existing technologies. Dynamic topology adjustment and priority preemption mechanisms work together: when a link failure triggers topology replanning, high-priority instructions can directly occupy 80% of the bandwidth resources of the new path through dedicated bandwidth tokens reserved on edge nodes (low-priority instructions are allocated only the remaining 20%). This resource tilting strategy ensures that emergency instructions can still be transmitted with low latency during network fluctuations. This mechanism is a unique design not disclosed in existing technologies.

[0050] 2. This invention assigns priorities to instructions through an intelligent hierarchical model. High-priority instructions can reserve resources in advance and be transmitted first. In case of emergency, resources can be preempted. The multi-path transmission strategy automatically switches when a certain path fails. Combined with fragment verification and blockchain records, data integrity is ensured. Edge-cloud redundant backup avoids instruction loss caused by single point of failure. These mechanisms together construct a multi-layered reliability assurance system, solving the problems of easy interference and insufficient security in instruction transmission in traditional methods.

[0051] 3. This invention uses intelligent priority grading to tilt resources toward high-priority instructions, avoiding resource waste. The collaborative scheduling mechanism of the edge node cluster can dynamically adjust resource allocation according to real-time load, while transmission quality feedback and adaptive optimization continuously optimize the strategy. This enables the system to efficiently process a large number of instructions even during peak flight take-off and landing periods, solving the problems of rigid resource allocation and poor adaptability of traditional methods, and improving the resource utilization of edge nodes and the system's ability to cope with complex scenarios.

[0052] 4. The reinforcement learning and priority model of this invention form a dynamic iteration: Reinforcement learning adjusts the feature weights of the MLP neural network through feedback data. For example, the weight of time-sensitive features is increased in emergency scenarios, and the weight of instruction type features is increased in non-emergency scenarios, making the priority classification more suitable for real-time transmission requirements. This closed-loop iteration of 'evaluation-transmission-optimization' is an innovative design that cannot be achieved by existing static classification models.

[0053] 5. The parameter correction logic of reinforcement learning and priority model in this invention is clearly defined: when a high-priority instruction delay exceeds the standard, the reinforcement learning output will increase the weight coefficient of the time-sensitive feature in the MLP neural network by 20%-30%; when the packet loss rate exceeds the standard, it will increase the weight of the link quality parameter Q in the path scoring model (increasing the β value by 0.1-0.2). This precise parameter mapping relationship allows the optimization process to directly serve the core requirements of aviation instructions, and no such linkage mechanism has been disclosed in existing technologies. Attached Figure Description

[0054] Figure 1 This is a flowchart of the decision command transmission method for control tower and flight support command room based on edge nodes according to the present invention. Detailed Implementation

[0055] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] In this invention, the selection of parameters (such as the number of hidden layer neurons in the MLP neural network and the size of the DQN experience replay pool) is determined based on the characteristics of aviation command scenarios and industry practices: the number of hidden layer neurons adopts a decreasing structure of 64, 32, and 16 to adapt to the balance between instruction feature dimensions and computational efficiency; the size of the experience replay pool is set according to the conventional configuration of reinforcement learning to ensure the stability of the optimization process.

[0057] like Figure 1 As shown, this embodiment of the invention provides a method for transmitting decision-making commands in the control tower and flight support command center based on edge nodes. The specific steps of this method are as follows:

[0058] S1. Edge Node Cluster Construction and Dynamic Networking: Deploy edge nodes containing NVIDIA Jetson chips, millimeter-wave communication modules and fiber optic interfaces in key areas of the airport. Build a distributed cluster through a self-organizing algorithm and use a sliding window algorithm to monitor link bandwidth, latency and packet loss rate in real time. When the indicators exceed the threshold, adjust the network topology using fiber optic backup links to establish a stable basic network for command transmission.

[0059] S2. Intelligent Priority Classification of Decision Instructions: Based on the established edge cluster, an MLP neural network model is run. Input features such as instruction type, flight status and time sensitivity are input. After nonlinear processing by the hidden layer, the output layer outputs four priority levels: emergency, high, medium and low through the softmax function.

[0060] S3. Priority-based edge node resource reservation: Based on the hierarchical results, the resource management module reserves CPU cores, contiguous memory space, and dedicated bandwidth tokens for high-priority instructions; in the event of an emergency instruction, a preemption mechanism is triggered to temporarily allocate low-priority resources so that critical instructions can obtain sufficient processing and transmission resources.

[0061] S4. Multi-path parallel transmission and dynamic route switching: After resource reservation is completed, medium and high priority instructions enable multi-path transmission. Paths are selected by hop count and link quality assessment. A monitoring thread tracks the status. When a path fails, a reselection algorithm switches to a high-quality path, so that instructions can flow efficiently in the cluster.

[0062] S5. Command content fragmentation and edge node collaborative verification: During transmission, large-capacity commands are fragmented according to their logical structure, and identifiers are generated using a hash algorithm and distributed to edge nodes. Nodes verify the integrity using a CRC algorithm, and the transmission information is recorded using a blockchain. After all fragments are received, the target node reassembles and performs a second verification to ensure the integrity of the command.

[0063] S6, Edge-Cloud Collaborative Command Redundancy Backup: After sharding verification, based on node load and network distance, command copies are synchronized to 2-3 nearby edge nodes, and erasure coding technology is used to prevent data loss; critical commands are synchronously uploaded to the cloud and long-term backup is achieved through object storage, establishing a two-layer security defense.

[0064] S7. Real-time feedback and adaptive optimization of transmission quality: Throughout the process, the target node calculates indicators such as latency and packet loss rate in real time and transmits them to the cluster via the feedback channel. The cluster initiates reinforcement learning based on the data, uses transmission performance as a reward signal, and dynamically adjusts the priority model, path strategy and resource allocation to continuously improve transmission quality.

[0065] In step S1, edge node cluster construction and dynamic networking refers to deploying edge nodes in the control tower, flight support command room, and key areas of the airport. The hardware uses NVIDIA Jetson series computing chips, whose powerful parallel computing capabilities can quickly process instruction data; it is equipped with millimeter-wave communication modules to achieve high-speed wireless connection, and is also equipped with fiber optic interfaces as redundancy backup. This configuration provides nodes with efficient processing and diverse communication capabilities, providing a hardware foundation for cluster construction.

[0066] Based on a self-organizing algorithm, nodes establish initial connections via millimeter waves to form a distributed cluster. A sliding window algorithm is used to monitor link quality in real time. When indicators such as bandwidth and latency exceed the limits, dynamic topology adjustment is initiated, utilizing fiber optic links to replan the path. The AODV protocol networking process is as follows: ① Nodes broadcast their own status information via Hello messages; ② The source node sends a route request (RREQ) to the target node, and intermediate nodes forward and record the route path; ③ Upon receiving the RREQ, the target node returns a route response (RREP), and the source node establishes a transmission path based on the RREP; ④ The sliding window algorithm samples the link status at fixed intervals. When the link quality does not meet the threshold, route rediscovery and path adjustment are triggered. This mechanism keeps the cluster network stable.

[0067] When the link delay exceeds 80% of Dmax (e.g., emergency command delay exceeds 40ms) or the packet loss rate exceeds 5%, dynamic topology adjustment is initiated, and the time threshold for fiber link replanning is set to 10ms.

[0068] The self-organizing algorithm adopts the AODV routing protocol in the AdHoc network, with a sliding window size of 100ms to adapt to the transmission cycle characteristics of aviation commands, ensuring the real-time and accuracy of link status monitoring, and sampling and updating the link bandwidth, latency, and packet loss rate every 50ms.

[0069] In step S2, the intelligent classification of decision command priority is based on the distributed computing power of the edge cluster. An MLP neural network priority evaluation model is built. The input layer integrates features such as command type, flight status and time sensitivity. For example, emergency avoidance command and flight association information during takeoff are included in the analysis to provide multi-dimensional data support for the model.

[0070] The model's hidden layers extract feature associations using non-linear activation functions, while the output layer uses a softmax function to generate a four-level priority probability distribution, automatically classifying instructions. This result accurately distinguishes the urgency of instructions, providing clear guidance for subsequent differentiated resource allocation and transmission strategies.

[0071] The MLP neural network contains three hidden layers with 64, 32, and 16 neurons respectively. The input features are normalized to the [0,1] interval using Min-Max. The hidden layers use the ReLU activation function, and the parameters of the softmax function in the output layer are set to default values.

[0072] Formula for calculating softmax priority probability:

[0073]

[0074] In the formula: P(y=k|z) is the probability that the instruction belongs to the k-th priority level; z kis the output value of the k-th neuron in the output layer of the neural network; C is the total number of priority levels (C=4, corresponding to urgent, high, medium, and low); e is a natural constant;

[0075] In step S3, priority-based edge node resource reservation refers to the resource management module starting a reservation mechanism based on the priority result, the CPU using a process scheduling algorithm to allocate independent cores for high-priority instructions, memory reserving continuous space through paging technology, and bandwidth allocating dedicated tokens through a token bucket algorithm to pre-configure resources for instructions of different priorities.

[0076] The process scheduling algorithm adopts preemptive priority scheduling, and the process priority value of high priority instructions is set to 1-4 (urgent instructions are 1, and the lowest is 4); the memory paging technology adopts contiguous physical page allocation, and the size of each page is 4KB; the token bucket algorithm has a bucket capacity of 1MB and a token generation rate of 500KB / s (the token bucket rate for high priority instructions is increased to 1MB / s).

[0077] The correspondence between process priority values ​​and CPU cores is as follows: priority 1 is bound to CPU0 core, priority 2 is bound to CPU1 core, priority 3 is bound to CPU2 core, and priority 4 is bound to CPU3 core, ensuring that high-priority instructions have exclusive access to dedicated core resources.

[0078] An emergency instruction (priority 1) triggers a CPU interrupt signal (IRQ7). Upon receiving the signal, low-priority processes (priority 3 / 4) immediately release the memory pages they occupy (marked as 'reclaimable' by the Memory Management Unit (MMU)). The bandwidth token bucket is reset by clearing the token counter (the high-priority token generation rate is temporarily increased to 2MB / s). The entire preemption process is completed within 5ms.

[0079] When an emergency command is detected, a preemption mechanism is triggered, which temporarily allocates some resources to low-priority commands based on a priority strategy to ensure that emergency commands can quickly obtain sufficient resources and avoid delays caused by insufficient resources.

[0080] The specific content of multi-path parallel transmission and dynamic route switching in step S4 is as follows: after completing resource reservation, multi-path transmission is started for medium and high priority instructions. Through the link quality comprehensive scoring model, probe packets are sent periodically to collect delay and bandwidth information to build a status table, thereby filtering multiple available paths. The multi-path design provides multiple channels for instruction transmission and improves the flexibility of transmission.

[0081] The expression for the comprehensive link quality scoring model is as follows:

[0082]

[0083] In the formula: S is the overall path score (the higher the value, the better the path); α and β are weight coefficients (α+β=1); H is the number of hops in the path; Q is the link quality parameter, a normalized value of the overall bandwidth, latency, and packet loss rate, ranging from [0,1]; when calculating the Q value, the weight allocation of bandwidth, latency, and packet loss rate is 4:4:2 (i.e., bandwidth accounts for 40%, latency for 40%, and packet loss rate for 20%), based on the equal priority requirement of low latency and high bandwidth in aviation scenarios;

[0084] The calculation rule for the path hop count H is as follows: Starting from the source node, the H value is incremented by 1 for each forwarding through an edge node (e.g., source node → node A → target node, H = 2), and H = 1 for a direct path.

[0085] A dedicated thread is set up to monitor the path status. When a path experiences increased latency or failure, a reselection algorithm is used to switch traffic to a high-quality path, achieving seamless connection of command transmission. This enables medium and high priority commands to flow efficiently in the cluster and reduces the risk of transmission interruption.

[0086] The reselection algorithm adopts an improved A* algorithm, using the path comprehensive score S as the heuristic function, and prioritizes the path with a better S value; the switching trigger condition is: the single path delay exceeds Dmax for 3 consecutive samples (interval of 50ms), or the packet loss rate exceeds 5% for 2 consecutive samples.

[0087] The improved A* algorithm introduces a link quality penalty factor: when the path packet loss rate is >3%, the heuristic function S increases by a penalty value of 0.2 to prioritize avoiding unstable paths; when the delay is >50% of Dmax, the penalty value increases to 0.5.

[0088] In the comprehensive link quality scoring model, α and β are dynamic weight coefficients (α+β=1): for emergency instructions, β=0.7 and α=0.3, and for non-emergency instructions, β=0.5 and α=0.5, based on the priority requirement of low latency in aviation scenarios.

[0089] In step S5, the instruction content fragmentation and edge node collaborative verification firstly involves fragmenting large-capacity instructions according to their logical structure and functional modules during multi-path transmission, generating unique identifiers using a hash algorithm, and allocating them to different edge nodes. Fragmentation reduces the pressure of single-path transmission, while the identifiers are used to achieve orderly data management.

[0090] Then, after each node receives the fragment, it uses the CRC algorithm to verify the integrity. Blockchain technology records the transmitted information to ensure that it cannot be tampered with. After all the fragments are received, the target node reassembles them according to the identifier and performs a second verification, thereby ensuring the integrity of the instruction data in all aspects and avoiding information loss or tampering during the transmission process.

[0091] The CRC algorithm uses the CRC32 standard to generate a 32-bit check value, which is transmitted along with the fragmented data. After the target node reassembles the data, it performs a second CRC check and compares it with the check value of the original instruction to ensure consistency.

[0092] Large-capacity instructions are fragmented according to the logical structure of 'instruction header-opcode-parameter-verification segment', with each fragment not exceeding 1MB in size; the hash algorithm uses SHA-256 to generate unique identifiers, and the blockchain record fields include fragment identifier, transmission node ID, and timestamp. It adopts a consortium blockchain architecture, and node consensus is achieved through the PBFT algorithm.

[0093] PBFT consensus mechanism steps: ① The master node sends a pre-preparation message (including shard hash and timestamp); ② After all nodes have verified the message, the master node sends a preparation message; ③ After collecting preparation messages from ≥2 / 3 of the nodes, the nodes send a commit message; ④ After receiving ≥2 / 3 of the commit messages, the shard information is written to the blockchain, and nodes that do not respond within the timeout period are marked as abnormal.

[0094] In step S6, the edge-cloud collaborative instruction redundancy backup refers to selecting 2-3 neighboring nodes to store instruction copies based on the DHT algorithm after fragment verification, and using erasure coding technology to encode redundant fragments. The DHT algorithm uses the Chord protocol and selects neighboring nodes based on the hash value of the node ID; the erasure coding uses RS(3,2) encoding, that is, one check block is generated for every two data blocks, and the original data can be restored if any two blocks are complete. By utilizing node collaboration, instruction can be quickly restored, the risk of single point of failure can be addressed, and the short-term security of data can be guaranteed.

[0095] For critical instructions, copies are simultaneously uploaded to the cloud and object storage technology is used for long-term preservation. Edge local backup and cloud backup work together to meet the needs of rapid recovery and achieve long-term data archiving, thus achieving multi-layered instruction security protection.

[0096] Object storage uses the S3 protocol (SimpleStorageService industry standard), with a storage format of JSON (metadata) + binary (instruction content), and reuses hash identifiers with transport fragments; the cloud fragment size is consistent with the edge transport fragment size (≤1MB), which facilitates cross-node verification;

[0097] In step S7, real-time transmission quality feedback and adaptive optimization refers to the target node's performance calculation module measuring indicators such as latency and packet loss rate in real time throughout the entire instruction transmission process, and feeding them back to the source node and cluster through a dedicated channel to provide the system with accurate transmission quality data.

[0098] Based on feedback data, the cluster uses reinforcement learning to build an optimization model. With transmission performance as the reward function, it adjusts variables such as priority models and path strategies. Through continuous iteration, it achieves dynamic optimization of strategies at each stage and continuously improves the overall transmission quality.

[0099] The DQN policy network adopts a 3-layer fully connected structure: the number of neurons in the input layer is consistent with the dimension of the state space (such as link delay, CPU load, and priority label, which together form 3 dimensions), the hidden layer has 64-32 neurons (ReLU activation function), and the number of neurons in the output layer is consistent with the dimension of the action space (such as adjusting β weights, modifying thresholds, and other 3-dimensional actions). The network minimizes the loss function through the Adam optimizer.

[0100] The DQN algorithm's experience replay pool size is 1000 samples (based on the command processing scale of the aviation command scenario, meeting the reinforcement learning's need for diverse historical data), and the learning rate is set to 0.001; Dmax is set according to command type: 50ms for emergency commands, 100ms for high-priority commands, and 200ms for medium / low-priority commands; Q-value normalization adopts the min-max method, and the calculation formula is Q = (measured value - minimum value) / (maximum value - minimum value), where the maximum bandwidth is 1Gbps, the maximum latency is 500ms, and the maximum packet loss rate is 10%;

[0101] The reinforcement learning uses the DQN algorithm, which updates the policy network after processing every 200 instructions. The optimization variables include the feature weights of the priority model (the weights of time-sensitive features are higher than those of instruction type features). In the reward function, k1, k2, and k3 are dynamic weights. For urgent instructions, k1 (delay weight) > k2 (packet loss rate weight) > k3 (integrity weight), and for non-urgent instructions, k3 > k1 = k2.

[0102] The specific weight values ​​are: emergency commands k1 = 0.6, k2 = 0.3, k3 = 0.1; non-emergency commands k1 = 0.2, k2 = 0.2, k3 = 0.6; the integrity index I is calculated as follows: I = number of received fragments / total number of fragments (if 1 fragment is lost, I = (n-1) / n, where n is the total number of fragments);

[0103] The state space includes: current link latency, node CPU load, and instruction priority label; the action space includes: adjusting the β weight in the path scoring model, modifying the priority grading feature threshold, and optimizing the resource reservation ratio. The DQN algorithm trains the state-action mapping relationship using link states and transmission quality samples from the experience replay pool.

[0104] Reinforcement learning reward function formula:

[0105]

[0106] In the formula: R is the transmission quality reward value; k1, k2, and k3 are weighting coefficients and the actual transmission delay; Dmax is the maximum allowable delay threshold; L is the packet loss rate; 1 is the integrity index, 1 = 1 for complete transmission, otherwise 1 < 1.

[0107] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0108] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for transmitting decision-making instructions in control towers and flight support command rooms based on edge nodes, characterized in that: The specific steps of this method are as follows: S1. Edge Node Cluster Construction and Dynamic Networking: Deploy edge nodes in key areas, build distributed clusters through self-organizing algorithms, and use sliding window algorithms to monitor link bandwidth, latency and packet loss rate in real time. S2. Intelligent Priority Classification of Decision Instructions: Based on the established edge cluster, the input instruction type, flight status and time sensitivity features are considered, and the output layer outputs four priority levels: emergency, high, medium and low through the softmax function. S3. Priority-based edge node resource reservation: Based on the classification results, the resource management module reserves CPU cores, contiguous memory space, and dedicated bandwidth tokens for high-priority instructions. S4. Multi-path parallel transmission and dynamic route switching: After resource reservation is completed, medium and high priority commands enable multi-path transmission, and the path is selected by hop count and link quality assessment. A monitoring thread is set to track the status. S5. Command content fragmentation and edge node collaborative verification: During transmission, large-capacity commands are fragmented according to logical structure, and identifiers are generated using a hash algorithm and distributed to edge nodes. The nodes verify the integrity through a CRC algorithm. S6, Edge-Cloud Collaborative Command Redundancy Backup: After sharding verification is completed, based on node load and network distance, command copies are synchronized to 2-3 nearby edge nodes, and erasure coding technology is used to prevent data loss; S7. Real-time feedback and adaptive optimization of transmission quality: Throughout the process, the target node calculates latency and packet loss rate in real time and transmits them to the cluster via the feedback channel. The cluster then initiates reinforcement learning based on the data.

2. The method for transmitting decision-making instructions in the control tower and flight support command room based on edge nodes according to claim 1, characterized in that: The specific steps for edge node cluster construction and dynamic networking in step S1 are as follows: S11. Node Deployment and Hardware Configuration: Edge nodes are deployed in the control tower, flight support command center, and key areas of the airport. The hardware uses NVIDIA Jetson series computing chips; millimeter-wave communication modules are equipped with fiber optic interfaces as redundancy backup. S12. Self-organizing networking and link optimization: Based on the self-organizing algorithm, nodes establish initial connections through millimeter waves to form a distributed cluster. The sliding window algorithm is used to monitor link quality in real time. When bandwidth and latency indicators exceed the standard, dynamic topology adjustment is initiated, and the optical fiber link is used to replan the path.

3. The method for transmitting decision-making instructions in the control tower and flight support command room based on edge nodes according to claim 1, characterized in that: The specific steps of intelligent priority grading of decision instructions in step S2 are as follows: S21. Evaluation Model Construction and Feature Input: Based on the distributed computing power of edge clusters, an MLP neural network priority evaluation model is built, and the input layer integrates features such as instruction type, flight status and time sensitivity. S22. Feature Processing and Priority Output: The hidden layer of the model extracts feature associations through a non-linear activation function, and the output layer uses the softmax function to generate a four-level priority probability distribution, automatically completing the instruction classification.

4. The method for transmitting decision-making instructions in the control tower and flight support command room based on edge nodes according to claim 1, characterized in that: The specific steps for priority-based edge node resource reservation in step S3 are as follows: S31. Conventional resource reservation mechanism: Based on the priority results, the resource management module starts the reservation mechanism. The CPU uses the process scheduling algorithm to allocate independent cores for high-priority instructions. The memory reserves continuous space through paging technology. The bandwidth is allocated with dedicated tokens through the token bucket algorithm. S32. Emergency Resource Scheduling Strategy: When an emergency command is detected, a preemption mechanism is triggered, and some resources of the low-priority command are temporarily allocated based on the priority strategy.

5. The method for transmitting decision-making instructions in the control tower and flight support command room based on edge nodes according to claim 1, characterized in that: The specific steps of multi-path parallel transmission and dynamic route switching in step S4 are as follows: S41. Multipath Probe and Selection: After resource reservation is completed, multipath transmission is initiated for medium and high priority commands. Through the link quality comprehensive scoring model, probe packets are sent periodically to collect latency and bandwidth information and build a status table. S42. Real-time monitoring and dynamic switching: A dedicated thread is set up to monitor the path status. When a path experiences increased latency or failure, the traffic is switched to a high-quality path through a reselection algorithm.

6. The method for transmitting decision-making instructions in the control tower and flight support command room based on edge nodes according to claim 1, characterized in that: The specific steps for instruction content fragmentation and edge node collaborative verification in step S5 are as follows: S51. Large-capacity instruction fragmentation: In multi-path transmission, large-capacity instructions are fragmented according to logical structure, split according to functional modules, and a unique identifier is generated by a hash algorithm and distributed to different edge nodes. S52. Fragmentation Verification and Reassembly Verification: After receiving the fragments, each node uses the CRC algorithm to verify their integrity. Blockchain technology records the transmitted information to ensure that it cannot be tampered with. After all fragments have arrived, the target node reassembles them according to their identifiers and performs a second verification.

7. The method for transmitting decision-making instructions in the control tower and flight support command room based on edge nodes according to claim 1, characterized in that: The specific steps for edge-cloud collaborative instruction redundancy backup in step S6 are as follows: S61. Edge node local redundant storage: After completing the fragmentation verification, select 2-3 neighboring nodes to store instruction copies based on the DHT algorithm, and use erasure coding technology to encode redundant fragments. S62, Cloud-based Long-Term Backup and Collaboration: For critical instructions, copies are synchronously uploaded to the cloud and object storage technology is used to achieve long-term preservation.

8. The method for transmitting decision-making instructions in the control tower and flight support command room based on edge nodes according to claim 1, characterized in that: The specific steps for real-time transmission quality feedback and adaptive optimization in step S7 are as follows: S71, Full-process performance monitoring and feedback: Throughout the entire command transmission process, the target node's performance calculation module measures latency and packet loss rate in real time and feeds them back to the source node and cluster through a dedicated channel; S72. Reinforcement Learning Dynamic Optimization: Based on feedback data, the cluster uses reinforcement learning to build an optimization model, using transmission performance as the reward function, and adjusts variables such as priority model and path strategy.

Images