An automatic order assignment method based on aircraft ground support at civil aviation airports

Abstract

The present invention relates to an automatic dispatching method based on aircraft ground support at civil aviation airports. By adopting a dynamic causality recognition module based on Bayesian network and combining it with a context awareness module of convolutional neural network (CNN) and long short-term memory network (LSTM), the prediction accuracy and task assignment efficiency are improved. The improved multi-objective optimization algorithm combines genetic algorithm and particle swarm optimization algorithm, and introduces a dynamic weight adjustment mechanism to realize the adaptive optimal allocation of resources. The path planning algorithm combining multi-level A* search and deep reinforcement learning can adjust the path in real time and dynamically avoid obstacles. The system enhances the real-time response ability and decision-making accuracy through a self-learning algorithm based on deep deterministic policy gradient (DDPG) and a comprehensive decision support system based on graph neural network (GNN). This method improves the operation efficiency, safety and emergency response ability of airport ground support services.

Description

Technical Field

[0001] The present invention belongs to the technical field of civil aviation, and specifically relates to an automatic dispatching method based on aircraft ground support at civil aviation airports. Background Art

[0002] With the rapid development of the civil aviation transportation industry, airport ground support services play a crucial role in ensuring flight punctuality, enhancing passenger satisfaction, and improving the overall operational efficiency of airports. To cope with the increasing ground support requirements, an automatic dispatching system has become one of the key technologies for improving ground service efficiency.

[0003] In the prior art, Chinese invention patent CN118134207B discloses a production scheduling method and system in the automatic allocation scenario of civil airport guiding vehicles. This method generates a more adaptable initial population by constructing an initial population set, obtaining and parsing current flight data to be allocated and guiding vehicle and personnel data, and using a genetic algorithm, thereby providing optimized genetic algorithm support for the allocation of airport guiding vehicles. However, this method has certain limitations in dealing with multi-source data fusion, real-time task adjustment, and path planning in complex environments, and it is difficult to fully cope with the dynamic changes and emergencies in airport operations.

[0004] Another Chinese invention patent CN118114952B discloses a rolling time-domain dynamic allocation method and system for civil airport boarding tasks. This method can achieve automatic rolling allocation of boarding tasks and improve the allocation efficiency and effect by initializing a task list, receiving and parsing historical operation data, constructing a dynamic allocation judgment structure, and adjusting the task allocation ratio according to the rolling time interval. However, this method mainly focuses on boarding task allocation, lacks comprehensive management of the entire airport ground support tasks, and has insufficient considerations in aspects such as data security, equipment maintenance management, and personnel training, and cannot comprehensively improve the overall efficiency and security of airport ground support services.

[0005] The above designs have improved the automation and allocation efficiency of ground support tasks by introducing genetic algorithms and rolling allocation mechanisms, but still have the following limitations: Insufficient multi-source data fusion: The existing methods lack advanced technical support in the integration of multi-source data and the identification of causal relationships, resulting in limited prediction accuracy and task allocation precision. Lack of real-time feedback and dynamic optimization: There is a lack of an effective real-time monitoring and feedback mechanism, making it difficult to adjust task allocation and resource configuration in a timely manner according to dynamic changes. Insufficient data security: During the transmission of multi-source data, there are no effective data security protection measures, and there is a risk of data tampering and unauthorized access. Imperfect equipment maintenance management: The ground support equipment is not maintained predictably, and it is easy to affect task execution due to equipment failures. Summary of the Invention

[0006] To solve the above problems, the present invention provides an automatic order dispatching method based on aircraft ground support at civil aviation airports, including the following steps:

[0007] S1. Use a time series prediction model to fuse multi-source data of historical flight data, weather forecasts, and air traffic control instructions, and utilize a dynamic causality recognition module based on a Bayesian network to identify the causal relationships between different data sources in real time to accurately predict the flight arrival time;

[0008] S2. Based on the prediction results of step S1, combined with real-time monitoring data, including flight dynamics, vehicle positions, and equipment status, utilize a context awareness module based on a convolutional neural network (CNN) and a long short-term memory network (LSTM) to dynamically adjust the priorities of tasks and perform dynamic task allocation;

[0009] S3. Apply an improved multi-objective optimization algorithm to adaptively optimize the allocation of resources according to factors such as the urgency of tasks and the types of required equipment. This optimization algorithm combines the global search ability of the genetic algorithm and the fast convergence characteristics of the particle swarm optimization algorithm, and introduces a dynamic weight adjustment mechanism and a real-time feedback loop, with self-learning ability, and can continuously optimize the decision-making logic based on historical data and real-time feedback;

[0010] S4. Use an improved path planning algorithm to plan the optimal path for each execution unit. The path planning algorithm is based on a model that combines multi-level A* search and deep reinforcement learning, can update in real time and adapt to the dynamic changes of the airport environment, and at the same time combines an energy management algorithm based on a prediction model to optimize energy consumption, ensuring that electric vehicles have enough energy to return to the base or reach the next charging point before completing the task;

[0011] S5. Through a real-time monitoring and feedback mechanism, establish a feedback loop, feed the real-time monitoring data back to the learning module of the system, and utilize a self-learning algorithm based on deep deterministic policy gradient (DDPG) and a comprehensive decision support system based on a graph neural network (GNN) to dynamically adjust task allocation and path planning, ensure the overall progress of the system in case of emergencies, and through an improved rapid response and resource allocation mechanism, including a priority-based dynamic resource allocation algorithm, a machine learning-based early warning model, a knowledge graph-based automated audit system, a collaborative filtering-based task allocation system, and an intelligent scheduling assistant based on augmented reality (AR) technology, quickly respond to emergencies and reasonably allocate resources;

[0012] S6. Adopt a secure data transmission module based on blockchain to ensure the integrity and security of data during the multi-source data fusion and real-time data transmission process, and prevent data tampering and unauthorized access;

[0013] S7. Integrate the predictive maintenance module, which predicts potential failures of ground support equipment based on real-time monitoring data and historical maintenance records, schedules maintenance tasks in advance, and reduces the impact of equipment failures on ground support tasks;

[0014] S8. Through the user behavior analysis module, analyze the operation habits and task execution efficiency of ground crew, optimize the task allocation strategy, and improve the overall efficiency of human-machine collaboration;

[0015] S9. Adopt an operator training module based on virtual reality (VR) and augmented reality (AR) technologies to provide an immersive training environment for ground crew, and improve their proficiency in operating the automatic dispatching system and emergency response capabilities;

[0016] S10. Utilize big data analysis and machine learning technologies to deeply mine historical task execution data, optimize the task allocation model and resource configuration strategy, and continuously improve the overall performance and service quality of the system.

[0017] The multi-source data fusion in step S1 includes time series data, meteorological image data, and air traffic control instruction text data, which are deeply integrated using a multi-modal data fusion algorithm based on deep learning, and the fusion ratio of different data sources is optimized using an adaptive weight allocation mechanism.

[0018] The situation awareness module in step S2 adopts a multi-modal data fusion algorithm based on convolutional neural network (CNN) and long short-term memory network (LSTM), and uses Internet of Things (IoT) technology and sensor networks to real-time acquire and integrate relevant data to accurately perceive the current operation status of the airport and changes in the external environment.

[0019] The improved multi-objective optimization algorithm in step S3 combines the global search ability of genetic algorithm and the fast convergence characteristics of particle swarm optimization algorithm, and introduces a dynamic weight adjustment mechanism, which improves the computational efficiency and solution accuracy of the optimization, and adapts to the requirements of complex and changeable ground support tasks.

[0020] The improved path planning algorithm in step S4 combines the heuristic search of multi-level A* search algorithm and the strategy optimization of deep reinforcement learning method, which can real-time plan and adjust the path in a complex airport environment, dynamically avoid static and dynamic obstacles, and improve the efficiency and accuracy of path planning through hierarchical path segment optimization.

[0021] The energy consumption optimization in step S4 further includes real-time monitoring of the battery status of electric vehicles, and dynamically adjusting the path planning using an energy management algorithm based on a prediction model to minimize energy consumption and charging times, and ensure that the vehicle has enough energy to return to the base or reach the next charging point before completing the task.

[0022] The improved rapid response and resource allocation mechanism in step S5 includes:

[0023] a. A hierarchical response system that classifies ground support tasks into different levels based on factors such as flight delay time, passenger demand, and weather conditions, and introduces a dynamic resource allocation algorithm based on priority;

[0024] b. A real-time monitoring and early warning platform that uses Internet of Things technology and sensor networks to monitor the status of all ground support vehicles and service facilities in real time, and identifies potential risks through an early warning model based on machine learning;

[0025] c. Standard operating procedures (SOPs) that establish detailed operation requirements, time limits, and quality standards for each type of ground support task, and ensure the strict implementation of SOPs through an automated audit system based on a knowledge graph;

[0026] d. Multidisciplinary collaboration teams (MDTs) composed of multiple relevant departments, responsible for handling major events or complex tasks, and optimizing team collaboration efficiency through a task allocation system based on collaborative filtering;

[0027] e. Automated and intelligent assistance tools, including intelligent dispatching assistant software and wearable devices based on augmented reality (AR) technology, which provide real-time guidance and task status display for ground crew.

[0028] The self-learning algorithm in step S5 is selected from a reinforcement learning algorithm based on deep deterministic policy gradient (DDPG), which can continuously optimize the scheduling strategy and path planning algorithm according to historical and real-time data, improving the intelligence level and response ability of the system.

[0029] The comprehensive decision support system integrates the information of modules for intelligent prediction, resource optimization, path planning, and real-time adjustment, adopts a unified decision-making framework based on graph neural network (GNN), and uses big data analysis and machine learning technologies to enhance the overall coordination of emergency response and the accuracy of decision-making.

[0030] The automatic order assignment system further includes a module for formulating standard operating procedures (SOPs) and training ground crew, ensuring that various ground support tasks are efficiently executed according to predetermined standards, and providing an immersive training environment for ground crew through an operator training module based on virtual reality (VR) and augmented reality (AR) technologies, improving their proficiency in operating the automatic order assignment system and emergency response ability.

[0031] In summary, the present application has the following beneficial effects:

[0032] 1. By adopting a time - series prediction model combined with a dynamic causality recognition module based on Bayesian networks, deep fusion and causal analysis are carried out on multi - source data such as historical flight data, weather forecasts, and air traffic control instructions, significantly improving the prediction accuracy of flight arrival times. The accurate prediction results provide a reliable data basis for subsequent task allocation, making task allocation more precise and efficient, and reducing resource waste and task delays caused by prediction errors.

[0033] 2. Through a context - awareness module based on convolutional neural networks (CNNs) and long short - term memory networks (LSTMs), the operating status of the airport and changes in the external environment can be sensed in real time. Combining real - time monitoring data, the system can dynamically adjust the priorities of tasks and perform intelligent task allocation to ensure that critical tasks are processed first, improving the overall work efficiency and service quality.

[0034] 3. Improve the multi - objective optimization algorithm, combine the advantages of genetic algorithms and particle swarm optimization algorithms, and introduce a dynamic weight adjustment mechanism and a real - time feedback loop to achieve adaptive optimal allocation of resources. This algorithm can efficiently allocate limited resources to meet the needs of complex and changing ground support tasks. At the same time, the improved path - planning algorithm combines multi - level A* search and deep reinforcement learning methods, can plan and adjust the optimal path of the execution unit in real time, dynamically avoid obstacles, and optimize the energy consumption of electric vehicles through an energy management algorithm. While ensuring the efficient completion of tasks, it reduces energy consumption and charging times and extends the service life of the vehicle.

[0035] 4. By introducing an improved rapid response and resource allocation mechanism, modules such as a hierarchical response system, a real - time monitoring and early warning platform, standardized operating procedures (SOPs), multi - disciplinary cooperation teams (MDTs), and automated and intelligent auxiliary tools are established, significantly improving the response speed of the system in emergencies and the rationality of resource allocation. The self - learning algorithm based on deep deterministic policy gradient (DDPG) and the integrated decision - making support system based on graph neural networks (GNNs) enable the system to continuously optimize scheduling strategies and path planning according to historical and real - time data, improving the accuracy of decision - making and overall coordination.

[0036] 5. Adopt a blockchain - based secure data transmission module to ensure data integrity and security during the process of multi - source data fusion and real - time data transmission, prevent data tampering and unauthorized access, and ensure the reliability of system operation. In addition, the integrated predictive maintenance module can predict potential equipment failures in advance based on real - time monitoring data and historical maintenance records, schedule maintenance tasks in a timely manner, reduce the impact of equipment failures on ground support tasks, and improve the reliability of equipment and the overall stability of the system.

[0037] 6. Through the user behavior analysis module, the operation habits and task execution efficiency of ground crew are deeply analyzed, the task allocation strategy is optimized, and the overall efficiency of human-machine collaboration is improved. At the same time, an operator training module based on virtual reality (VR) and augmented reality (AR) technologies is adopted to provide an immersive training environment for ground crew, enhancing their proficiency in operating the automatic order dispatching system and emergency response capabilities, ensuring that ground crew can execute various ground support tasks efficiently and accurately, and further improving the user experience and service quality.

[0038] 7. Through big data analysis and machine learning technologies, historical task execution data is deeply mined and analyzed to continuously optimize the task allocation model and resource configuration strategy, ensuring that the system can continuously adapt to the changing airport operation requirements and continuously improve the overall performance and service quality. The system's self-learning ability and feedback optimization mechanism endow it with the ability of self-improvement and self-optimization, enabling it to maintain an efficient and intelligent operation state during long-term operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is the work flow of this application Figure 1 ;

[0040] Figure 2 is the work flow of this application Figure 2 . DETAILED DESCRIPTION OF THE EMBODIMENTS

[0041] In order to make the objectives, technical solutions and beneficial effects of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings for the convenience of those skilled in the art to understand.

[0042] Embodiment 1:

[0043] As Figures 1 to 2 shown, this embodiment details an automatic order dispatching method for aircraft ground support at civil aviation airports. The core automatic order dispatching process covers key technical steps such as time series prediction, multi-source data fusion, causal inference, context awareness, multi-objective optimization, and path planning. Through this method, efficient and intelligent allocation and execution of ground support tasks at civil aviation airports can be achieved, ensuring the smooth and efficient operation of airport operations.

[0044] Step S1: Multi-source data fusion and flight arrival time prediction

[0045] In the initial stage of the automatic task assignment for ground support, the system first needs to accurately predict the arrival time of the upcoming flights. To this end, a time series prediction model is used to deeply fuse multi-source data such as historical flight data, weather forecasts, and air traffic control instructions. Specifically, the multi-source data fusion includes time series data, meteorological image data, and air traffic control instruction text data, and a multi-modal data fusion algorithm based on deep learning is used for deep integration. To further improve the prediction accuracy, the system introduces a dynamic causality recognition module based on Bayesian network to identify the causal relationships between different data sources in real time, and uses an adaptive weight allocation mechanism to optimize the fusion ratio of different data sources. Through this method, the system can accurately predict the arrival time of flights and provide a reliable data basis for subsequent task assignments.

[0046] In the initial stage of the automatic task assignment for ground support, the system first needs to accurately predict the arrival time of the upcoming flights. To this end, a time series prediction model is used to deeply fuse multi-source data such as historical flight data, weather forecasts, and air traffic control instructions.

[0047] Multi-source data fusion

[0048] The multi-source data fusion includes time series data, meteorological image data, and air traffic control instruction text data. Let:

[0049] · The time series data set is T = {T1, T2,..., T n}, where T i represents the historical data of the i-th flight.

[0050] · The meteorological image data set is I = {I1, I2,..., I m}, where I j represents the meteorological image of the j-th day.

[0051] · The air traffic control instruction text data set is C = {C1, C2,..., C k}, where C l represents the air traffic instruction of the l-th flight.

[0052] A multi-modal data fusion algorithm based on deep learning is used to deeply integrate different types of data. The specific fusion process is as follows:

[0053] F = Fusion(T, I, C)

[0054] where Fusion(·) represents the multi-modal data fusion function, which generates the fused feature vector F.

[0055] Dynamic causality recognition based on Bayesian network

[0056] To further improve the accuracy of prediction, the system introduces a dynamic causality recognition module based on Bayesian network to identify the causal relationships between different data sources in real time. A Bayesian network is a directed acyclic graph (DAG) used to represent the conditional dependence relationships between variables, and its structure can be expressed as:

[0057] G = (V, E)

[0058] where V represents the set of variables, E represents the set of directed edges, indicating the causal relationships between variables.

[0059] Through the Bayesian network, the conditional probability distribution P(X i ∣Parents(X i )) is defined for each variable, where Parents(X i ) represents the set of parent nodes of node X i . The dynamic causality recognition process includes:

[0060] Structure learning: According to historical data, learn the structure of the Bayesian network through algorithms (such as greedy search, genetic algorithm, etc.).

[0061] Parameter learning: After determining the network structure, use the maximum likelihood estimation or Bayesian estimation method to learn the conditional probability distribution of each node.

[0062] Finally, the Bayesian network is used to identify the causal relationships in multi-source data, thereby enhancing the interpretability and accuracy of the prediction model.

[0063] Adaptive weight allocation mechanism

[0064] To optimize the fusion ratio of different data sources, an adaptive weight allocation mechanism is introduced. Set the weights of each modality data in the fused feature vector F as w = [w T , w I , w C , where:

[0065] w T + w I + w C = 1

[0066] The weight allocation mechanism dynamically adjusts the weights through the following formula:

[0067]

[0068] where α m is a learnable parameter, which is optimized and adjusted according to the prediction error through the backpropagation algorithm.

[0069] Class arrival time prediction model

[0070] After completing multi-source data fusion and causal relationship identification, the system uses a time series prediction model to process the fused feature vectors and predict the flight arrival time. The prediction model is set as a Long Short-Term Memory network (LSTM) or Transformer, and its prediction process can be expressed as:

[0071]

[0072] Among them, is the predicted flight arrival time, and θ is the model parameter, which is optimized through the training process.

[0073] Working process

[0074] Data collection and fusion: The system uses Internet of Things (IoT) technology and sensor networks to collect multi-source data such as historical flight data, weather forecasts, and air traffic control instructions in real time. Using a multi-modal data fusion algorithm based on deep learning, time series data, meteorological image data, and air traffic control instruction text data are deeply integrated to generate the fused feature vector F.

[0075] Causal relationship identification: Through a dynamic causal relationship identification module based on a Bayesian network, the causal relationship map between multi-source data is identified and updated in real time to ensure the relevance of the fused data and the accuracy of the prediction model.

[0076] Weight optimization: The adaptive weight allocation mechanism dynamically adjusts the weights w of each modal data according to the real-time data and the results of causal relationship analysis, optimizes the data fusion ratio, and improves the performance of the prediction model.

[0077] Time series prediction: Use an LSTM or Transformer model to process the fused feature vector F and predict the flight arrival time Provide a reliable data basis for subsequent task allocation.

[0078] Step S2: Situation awareness and dynamic task allocation

[0079] Based on the flight arrival time prediction result of step S1, the system combines real-time monitoring data, including flight dynamics, vehicle positions, equipment status, etc., and uses a situation awareness module based on a Convolutional Neural Network (CNN) and a Long Short-Term Memory network (LSTM) to accurately perceive the current airport operation status and external environment changes. Specifically, the situation awareness module uses a multi-modal data fusion algorithm to use IoT technology and sensor networks to obtain and integrate relevant data in real time, thereby dynamically adjusting the priority of tasks and performing real-time task allocation. In this way, the system can flexibly adjust resource allocation according to the actual situation, ensure that key tasks are given priority, and improve the overall work efficiency.

[0080] Real-time Data Collection and Preprocessing

[0081] The system collects the following data in real time through Internet of Things (IoT) technology and sensor networks:

[0082] · Flight dynamic data D f ={d f1 , d f2 ,..., d fn}

[0083] · Vehicle location data D v ={d v1 , d v2 ,..., d vm}

[0084] · Equipment status data D e ={d e1 , d e2 ,..., d ek}

[0085] These data are first preprocessed, including steps such as data cleaning and normalization, to ensure data quality and consistency.

[0086] Multi-modal Data Fusion

[0087] The context awareness module uses a multi-modal data fusion algorithm to integrate different types of data into a unified feature representation:

[0088] F s = Fusion(D f , D v , D e )

[0089] where Fusion(·) represents the multi-modal data fusion function, generating the fused feature vector F s .

[0090] Convolutional Neural Network (CNN) Feature Extraction

[0091] For flight dynamic data and equipment status data, CNN is used for feature extraction:

[0092] F cnn = CNN(D f , D e ; θ cnn )

[0093] where θ cnn are the parameters of CNN, and F cnn is the extracted feature vector.

[0094] Long Short-Term Memory Network (LSTM) Sequence Modeling

[0095] For vehicle position data, LSTM is used for sequence modeling:

[0096] F lstm = LSTM(D v ; θ lstm )

[0097] where θ lstm is the parameter of LSTM, and F lstm is the extracted time series feature vector.

[0098] Situation Feature Fusion and Perception

[0099] Fuse the feature vectors extracted by CNN and LSTM to generate a comprehensive situation feature vector:

[0100] F context = Concat(F cnn , F lstm )

[0101] where Concat(·) represents the concatenation operation of feature vectors.

[0102] Task Priority Adjustment

[0103] Based on the comprehensive situation feature, the system calculates the priority score of each task through a fully connected layer:

[0104] p i = σ(W · F context + b)

[0105] where p i is the priority score of the i-th task, W and b are the weights and biases of the fully connected layer, and σ(·) is the activation function (such as the Sigmoid function).

[0106] Dynamic Task Allocation

[0107] According to the calculated task priority scores, the system adopts a sorting and allocation algorithm to allocate tasks to the most suitable resources (such as guiding vehicles, personnel, equipment):

[0108]

[0109] where T i represents the i-th task, R j represents the j-th resource, and Compatibility(·) represents the compatibility scoring function between the task and the resource.

[0110] Step S3: Multi-Objective Optimization and Resource Allocation

[0111] After the task allocation is completed, the system needs to optimize the allocation of available resources to meet the requirements of different tasks. To this end, an improved multi-objective optimization algorithm is adopted, which combines the global search ability of the genetic algorithm and the fast convergence characteristics of the particle swarm optimization algorithm, and introduces a dynamic weight adjustment mechanism and a real-time feedback loop. This optimization algorithm has the ability of self-learning and can continuously optimize the decision-making logic according to historical data and real-time feedback. By comprehensively considering factors such as the urgency of tasks and the types of required equipment, the system realizes the adaptive optimization allocation of resources, ensuring the maximization of resource utilization and the high efficiency of task execution.

[0112] After the task allocation is completed, the system needs to optimize the allocation of available resources to meet the requirements of different tasks. To this end, an improved multi-objective optimization algorithm is adopted, which combines the global search ability of the genetic algorithm (Genetic Algorithm, GA) and the fast convergence characteristics of the particle swarm optimization algorithm (Particle Swarm Optimization, PSO), and introduces a dynamic weight adjustment mechanism and a real-time feedback loop. This optimization algorithm has the ability of self-learning and can continuously optimize the decision-making logic according to historical data and real-time feedback. By comprehensively considering factors such as the urgency of tasks and the types of required equipment, the system realizes the adaptive optimization allocation of resources, ensuring the maximization of resource utilization and the high efficiency of task execution.

[0113] Multi-objective optimization problems usually involve multiple conflicting objective functions. In this invention, the following two objectives are mainly considered:

[0114] Minimize the resource usage cost C and maximize the task completion efficiency E. The mathematical expressions are as follows:

[0115]

[0116] Where:

[0117] · N is the total number of resources, and M is the total number of tasks.

[0118] · c i is the usage cost of the i-th resource.

[0119] · x i is the allocation status of resource i (0 means not allocated, 1 means allocated).

[0120] · e j is the completion efficiency of the j-th task.

[0121] · y j is the completion status of task j (0 means not completed, 1 means completed).

[0122] Combination of Genetic Algorithm and Particle Swarm Optimization Algorithm

[0123] To combine the global search ability of GA and the fast convergence characteristics of PSO, a hybrid optimization algorithm (Hybrid GA-PSO) is designed. The specific steps are as follows:

[0124] · Initialize the population: Generate the initial population P = {p1, p2,..., p G}, where each individual p g represents a resource allocation scheme.

[0125] · Evaluate the fitness: Calculate the fitness value F(p g ) of each individual, based on the objective functions C and E.

[0126] · Selection operation: Use roulette wheel selection or tournament selection to select individuals with higher fitness from the population.

[0127] · Crossover and mutation: Perform crossover and mutation operations on the selected individuals to generate a new generation of population.

[0128] · Particle swarm optimization: Use the new generation of population generated by GA as the initial particle swarm of PSO, and further optimize the resource allocation by updating the velocity and position of the particles.

[0129]

[0130] Where:

[0131] · is the velocity of particle i at the k-th generation.

[0132] · is the position of particle i at the k-th generation.

[0133] · w is the inertia weight.

[0134] · c1, c2 are acceleration constants.

[0135] · r1, r2 are random numbers.

[0136] · pbest i is the best position of particle i so far.

[0137] · gbest is the global best position.

[0138] Dynamic weight adjustment mechanism

[0139] To adapt to the dynamic changes of different tasks and resources, the system introduces a dynamic weight adjustment mechanism. The weights α and β are dynamically adjusted to balance the cost and efficiency objectives:

[0140] α(t + 1) = α(t) + Δα

[0141] β(t + 1) = β(t) - Δβ

[0142] Among them, Δα and Δβ are adjusted according to real-time feedback to ensure the target balance during the optimization process.

[0143] Real-time feedback loop

[0144] The system monitors the task execution in real time, feeds the actual execution data back to the optimization algorithm, and adjusts the optimization strategy in real time. The specific steps include:

[0145] · Monitor execution data: Collect the real-time data D of task execution exec 。

[0146] · Update optimization parameters: Adjust the optimization parameter θ according to the execution data.

[0147] · Re-optimize resource allocation: Based on the latest parameters and data, re-run the optimization algorithm and update the resource allocation plan.

[0148] θ new = θ old + Δθ

[0149] Among them, Δθ is calculated from the feedback data and is used to adjust the running parameters of the optimization algorithm.

[0150] Step S4: Path planning and energy consumption optimization

[0151] To ensure that each execution unit can complete the task efficiently, the system adopts an improved path planning algorithm to plan the optimal path for each execution unit. The path planning algorithm is based on a model that combines multi-level A* search and deep reinforcement learning, and can plan and adjust the path in real time in a complex airport environment, dynamically avoiding static and dynamic obstacles. In addition, combined with an energy management algorithm based on a prediction model, the system optimizes the energy consumption of electric vehicles, monitors the battery status in real time, and dynamically adjusts the path planning to minimize energy consumption and the number of charging times, ensuring that the vehicle has enough energy to return to the base or reach the next charging point before completing the task.

[0152] After the task allocation is completed, the system needs to optimize the allocation of available resources to meet the requirements of different tasks. To this end, an improved multi-objective optimization algorithm is adopted, which combines the global search ability of the Genetic Algorithm (GA) and the fast convergence characteristics of the Particle Swarm Optimization (PSO), and introduces a dynamic weight adjustment mechanism and a real-time feedback loop. This optimization algorithm has the ability of self-learning and can continuously optimize the decision-making logic according to historical data and real-time feedback. By comprehensively considering factors such as the urgency of tasks and the types of required equipment, the system realizes the adaptive optimization allocation of resources, ensuring the maximization of resource utilization and the efficiency of task execution.

[0153] Multi-objective optimization problems usually involve multiple conflicting objective functions. In this invention, the following two objectives are mainly considered:

[0154] Minimize the resource usage cost C and maximize the task completion efficiency E. The mathematical expressions are as follows:

[0155]

[0156] Where:

[0157] · N is the total number of resources, and M is the total number of tasks.

[0158] · c i is the usage cost of the i-th resource.

[0159] · x i is the allocation status of resource i (0 means not allocated, 1 means allocated).

[0160] · e j is the completion efficiency of the j-th task.

[0161] · y j is the completion status of task j (0 means not completed, 1 means completed).

[0162] Combination of Genetic Algorithm and Particle Swarm Optimization Algorithm

[0163] In order to combine the global search ability of GA and the fast convergence characteristics of PSO, a hybrid optimization algorithm (Hybrid GA-PSO) is designed. The specific steps are as follows:

[0164] · Initialize the population: Generate the initial population P = {p1, p2,..., p G}, where each individual p g represents a resource allocation scheme.

[0165] · Evaluate the fitness: Calculate the fitness value F(p g),based on the objective functions C and E.

[0166] · Selection operation: Use roulette wheel selection or tournament selection to select individuals with higher fitness from the population.

[0167] · Crossover and mutation: Perform crossover and mutation operations on the selected individuals to generate a new generation of population.

[0168] · Particle swarm optimization: Use the new generation of population generated by GA as the initial particle swarm of PSO, and further optimize the resource allocation by updating the velocity and position of the particles.

[0169] The particle swarm optimization update formula is as follows:

[0170]

[0171] Where:

[0172] · is the velocity of particle i in the k-th generation.

[0173] · is the position of particle i in the k-th generation.

[0174] · w is the inertia weight.

[0175] · c1 and c2 are acceleration constants.

[0176] · r1 and r2 are random numbers.

[0177] · pbest i is the best position of particle i so far.

[0178] · gbest is the global best position.

[0179] Dynamic weight adjustment mechanism

[0180] To adapt to the dynamic changes of different tasks and resources, the system introduces a dynamic weight adjustment mechanism. The weights α and β are dynamically adjusted to balance the cost and efficiency objectives:

[0181] α(t + 1) = α(t) + Δα

[0182] β(t + 1) = β(t) - Δβ

[0183] Where, Δα and Δβ are adjusted according to real-time feedback to ensure the objective balance in the optimization process.

[0184] Real-time feedback loop

[0185] The system monitors the task execution situation in real time, feeds the actual execution data back to the optimization algorithm, and adjusts the optimization strategy in real time. The specific steps include:

[0186] · Monitor execution data: Collect real-time data D of task execution exec 。

[0187] · Update optimization parameters: Adjust the optimization parameter θ according to the execution data

[0188] · Re-optimize resource allocation: Based on the latest parameters and data, re-run the optimization algorithm and update the resource allocation plan

[0189] θ new = θ old + Δθ

[0190] where Δθ is calculated from the feedback data and is used to adjust the running parameters of the optimization algorithm

[0191] Step S5: Real-time monitoring and feedback optimization

[0192] During the task execution, the system establishes a feedback loop through a real-time monitoring and feedback mechanism, and feeds the real-time monitoring data back to the learning module of the system. Using a self-learning algorithm based on Deep Deterministic Policy Gradient (DDPG) and a comprehensive decision support system based on Graph Neural Network (GNN), the system can dynamically adjust task allocation and path planning to ensure the stability of the overall progress in case of emergencies. At the same time, the system improves the rapid response and resource allocation mechanism, including a dynamic resource allocation algorithm based on priority, a warning model based on machine learning, an automated audit system based on knowledge graph, a task allocation system based on collaborative filtering, and an intelligent scheduling assistant based on Augmented Reality (AR) technology, to quickly respond to emergencies and rationally allocate resources, improving the overall response speed and decision-making accuracy of the system

[0193] Establish a feedback loop

[0194] The system collects real-time data D by monitoring the task execution in real time monitor , and feeds it back to the learning module

[0195] D feedback = D monitor

[0196] Self-learning algorithm (DDPG)

[0197] Use the DDPG algorithm to optimize the scheduling strategy and path planning. DDPG consists of an actor network μ(s|θ μ ) and a critic network

[0198] Q(s, a|θQ ) consists of, where:

[0199]

[0200] where s t is the state, a t is the action, r t is the reward, γ is the discount factor, is the noise, θ μ and θ Q are the parameters of the actor and critic networks respectively.

[0201] Integrated Decision Support System (GNN)

[0202] The decision support system based on GNN processes multi-source data through the graph structure G=(V, E), where the nodes V represent tasks and resources, and the edges E represent the relationships between them:

[0203] H (l+1) =σ(D -1 / 2 AD -1 / 2 H (l) w (l) )

[0204] where A is the adjacency matrix, D is the degree matrix, H (l) is the node feature of the l-th layer, W (l) is the weight matrix of the l-th layer, and σ is the activation function.

[0205] Resource Dynamic Allocation Algorithm

[0206] The resource dynamic allocation based on priority optimizes the objective function:

[0207]

[0208] where p i is the priority of task i, and x i is the resource allocation decision variable (0 or 1).

[0209] Early Warning Model

[0210] The early warning model based on machine learning uses classification algorithms to predict potential risks:

[0211]

[0212] where is the risk prediction result.

[0213] Automated Audit System

[0214] The automated audit system based on the knowledge graph verifies the compliance of task allocation with operating procedures:

[0215]

[0216] Step S6: Secure data transmission

[0217] To ensure data integrity and security during multi-source data fusion and real-time data transmission, the system adopts a secure data transmission module based on blockchain. This module utilizes the immutability and distributed storage characteristics of blockchain technology to prevent data from being tampered with and unauthorized access during transmission, ensuring the security and reliability of system data.

[0218] Blockchain network construction

[0219] Construct a private blockchain network where N represents the node set and E represents the connections between nodes.

[0220]

[0221] Data sharding and hashing

[0222] Shard the multi-source data D = {D1, D2,..., D n}, and perform hashing on each data shard:

[0223] H(D i ) = Hash(D i )

[0224] where H(D i ) is the hash value of the data shard D i .

[0225] Block generation and chain linking

[0226] Pack the hash values H(D i ) into blocks B k in chronological order, and ensure the immutability of data through chain linking:

[0227] B k = {H(D1), H(D2),..., H(D k ), H(B k-1 )}

[0228] where H(B k-1 ) is the hash value of the previous block.

[0229] Step S7: Predictive maintenance

[0230] The system integrates a predictive maintenance module that predicts potential failures of ground support equipment based on real-time monitoring data and historical maintenance records. By scheduling maintenance tasks in advance, the impact of equipment failures on ground support tasks is reduced, ensuring the efficient operation of the equipment and the continuity of airport ground services.

[0231] Data collection and preprocessing

[0232] Collect real-time monitoring data D of ground support equipment real ={d real1 , d real2 ,..., d realn} and historical maintenance records D hist ={d hist1 , d hist2 ,..., d histm}, and perform data cleaning and normalization:

[0233] D clean =Preprocess(D real , D hist )

[0234] Fault prediction model

[0235] Build a machine learning-based fault prediction model PredictFault(D clean ; θ pf ), and predict the equipment failure probability P(F j ):

[0236] P(F j ) = PredictFault(D clean，j ; θ pf )

[0237] where F j represents the fault event of the jth device, and θ pf is the model parameter.

[0238] Maintenance task scheduling

[0239] According to the fault prediction results, schedule maintenance tasks M j in advance:

[0240]

[0241] where τ is the fault probability threshold.

[0242] Optimize maintenance resource allocation

[0243] Adopt an optimization algorithm to allocate maintenance resources, minimize the maintenance cost C m and maximize the equipment availability U e :

[0244]

[0245] Among them, c j is the cost of the jth maintenance task, and m j is the allocation status of the maintenance task (0 or 1), and u j is the value of equipment availability improvement.

[0246] Step S8: User behavior analysis and optimization

[0247] Through the user behavior analysis module, the system analyzes the operation habits and task execution efficiency of ground crew, optimizes the task allocation strategy, and improves the overall efficiency of human-machine collaboration. This module uses big data analysis and machine learning technologies to mine user behavior patterns and provide a more accurate basis for task allocation.

[0248] Behavior data collection

[0249] Collect the operation data D user ={d u1 , d u2 ,..., d un} of ground crew, including information such as task completion time, operation steps, resource usage:

[0250] D user ={d u1 , d u2 ,..., d un}

[0251] Feature extraction and representation

[0252] Extract features from the collected user behavior data to form a user behavior feature vector F user :

[0253] F user =FeatureExtraction(D user )

[0254] Behavior pattern recognition

[0255] Use a clustering algorithm (such as K-Means) to identify the operation habits and behavior patterns of ground crew:

[0256] C = Cluster(F user ; K)

[0257] Among them, C is the clustering result and K is the number of clusters.

[0258] Task allocation optimization

[0259] Optimize the task assignment strategy based on the recognized behavior patterns, and assign tasks through a weighted scoring mechanism:

[0260] Score(T i , U j ) = α·Efficiency(U j ) + β·Preference(U j )

[0261] where T i is the task, U j is the ground crew, and α and β are weight coefficients.

[0262] Application of machine learning model

[0263] Apply a supervised learning model (such as random forest) to predict the adaptability of the ground crew to tasks:

[0264]

[0265] where is the adaptability score of the ground crew U j to the task T i , and θ ml is the model parameter.

[0266] Step S9: Operator training

[0267] The system adopts an operator training module based on virtual reality (VR) and augmented reality (AR) technologies to provide an immersive training environment for the ground crew. Through virtual scene simulation and real-time guidance, improve the ground crew's proficiency in operating the automatic dispatching system and emergency response capabilities, and ensure that they can efficiently handle various complex situations in actual operations.

[0268] Step S10: System continuous optimization

[0269] Using big data analysis and machine learning technologies, the system deeply mines the historical task execution data, optimizes the task assignment model and resource allocation strategy, and continuously improves the overall performance and service quality of the system. Through continuous data accumulation and analysis, the system can self-improve and adapt to the changing airport operation requirements.

[0270] Data collection and storage

[0271] The system continuously collects historical task execution data D history = {D1, D2,..., D n}, including task types, execution times, resource usage, etc., and stores it in the big data platform:

[0272] D history = {D1, D2,..., Dn}

[0273] Data Preprocessing and Feature Engineering

[0274] Clean, normalize historical data, and extract key feature F history :

[0275] F history = FeatureExtraction(D history )

[0276] Machine Learning Model Training

[0277] Use supervised learning algorithms (such as random forest, gradient boosting tree) to train the task assignment model Model TA and the resource allocation model Model RC :

[0278] Model TA = Train(F history ; θ TA )

[0279] Model RC = Train(F history ; θ RC )

[0280] where θ TA and θ RC are the parameters of the task assignment and resource allocation models respectively.

[0281] Model Evaluation and Optimization

[0282] Evaluate the model performance through cross-validation and performance metrics (such as accuracy, F1-score), and perform parameter tuning:

[0283]

[0284] Working Process

[0285] Data Collection and Fusion: The system uses Internet of Things (IoT) technology and sensor networks to collect multi-source data such as historical flight data, weather forecasts, and air traffic control instructions in real time. Using the multi-modal data fusion algorithm of deep learning, combined with the dynamic causal relationship recognition module based on Bayesian networks, different data sources are deeply integrated, and the data fusion ratio is optimized to ensure the comprehensiveness and accuracy of the data.

[0286] Flight Arrival Time Prediction: Based on the fused multi-source data, the system uses a time series prediction model, combined with causal inference, to accurately predict the arrival time of flights, providing basic data for subsequent task assignment.

[0287] Situation Awareness and Task Allocation: Combining the prediction results and real-time monitoring data, the situation awareness module uses a multimodal data fusion algorithm of CNN and LSTM to accurately perceive the current operating status of the airport and changes in the external environment, dynamically adjust task priorities, and perform task allocation to ensure that critical tasks are processed first.

[0288] Optimal Resource Allocation: Through an improved multi-objective optimization algorithm, the system adaptively optimizes the allocation of resources according to the urgency of tasks and equipment requirements, improving resource utilization efficiency and task execution accuracy.

[0289] Path Planning and Energy Management: Using an improved path planning algorithm, the optimal path is planned for each execution unit and adjusted in real time to adapt to the dynamically changing airport environment. At the same time, combined with an energy management algorithm, the energy consumption of electric vehicles is optimized to ensure that the vehicle has enough energy to return to the base or reach the next charging point while the task is completed.

[0290] Real-time Monitoring and Feedback Optimization: Through the real-time monitoring and feedback mechanism, the system uses the DDPG self-learning algorithm and the GNN integrated decision support system to dynamically adjust task allocation and path planning to ensure that the overall progress of the system is not affected in case of emergencies, and through an improved rapid response and resource allocation mechanism, it quickly responds and rationally allocates resources.

[0291] Data Security and Maintenance Management: Blockchain technology is used to ensure the security of data transmission, and a predictive maintenance module is integrated to predict potential equipment failures in advance, schedule maintenance tasks, and reduce the impact of equipment failures on ground support.

[0292] User Behavior Optimization and Training: Through the user behavior analysis module, the task allocation strategy is optimized to improve the efficiency of human-machine collaboration; an operator training module using VR and AR technologies is adopted to improve the operation proficiency and emergency response ability of ground crew.

[0293] System Continuous Optimization: Using big data analysis and machine learning technologies, the historical task execution data is deeply mined to optimize the task allocation model and resource allocation strategy, continuously improving the system performance and service quality.

[0294] Multi-source data fusion and causal relationship identification based on Bayesian networks improve the accuracy of flight arrival time prediction, providing a reliable data basis for subsequent task allocation.

[0295] The situation awareness module based on CNN and LSTM can accurately perceive the real-time operating status of the airport and changes in the external environment, dynamically adjust task priorities, and ensure that critical tasks are given priority.

[0296] The improved multi-objective optimization algorithm combines the advantages of genetic algorithm and particle swarm optimization algorithm, enhances the computational efficiency of resource allocation and the accuracy of solutions, and adapts to the requirements of complex and changeable ground support tasks.

[0297] The path planning algorithm combining multi-level A* search and deep reinforcement learning can adjust the path in real time, avoid obstacles, and optimize the energy consumption of electric vehicles through the energy management algorithm to ensure the efficient completion of tasks.

[0298] Through the DDPG self-learning algorithm and the GNN comprehensive decision support system, the system can dynamically adjust task allocation and path planning according to real-time data, ensure that the overall progress of the system is not affected in case of emergencies, and enhance the emergency response ability.

[0299] The security data transmission module based on blockchain ensures the integrity and security of data transmission; the predictive maintenance module predicts equipment failures in advance, reduces the impact of equipment failures on ground support tasks, and improves the reliability of the system.

[0300] The user behavior analysis module optimizes the task allocation strategy and improves the efficiency of human-machine collaboration; the operator training module based on VR and AR technologies enhances the operation proficiency and emergency response ability of ground crew, ensuring the efficient operation of the system.

[0301] Through big data analysis and machine learning technologies, the system can continuously optimize the task allocation model and resource allocation strategy, improve the overall performance and service quality, and ensure that the system can adapt to the changing airport operation requirements.

[0302] In summary, this embodiment provides a comprehensive, intelligent, and efficient automatic order assignment method for aircraft ground support in civil aviation airports, significantly improving the overall operation efficiency, safety, and emergency response ability of airport ground support services, and meeting the requirements of high-efficiency and intelligent ground support in modern civil aviation airports.

[0303] Embodiment 2:

[0304] As Figures 1 to 2 shown, this embodiment details the method of further enhancing the security and maintenance management functions of the system on the basis of Embodiment 1. By introducing an improved rapid response and resource allocation mechanism, self-learning algorithm, comprehensive decision support system, and standardized operation procedures and ground crew training module, the system can respond quickly in case of emergencies, allocate resources reasonably, and ensure the reliability of equipment and the efficiency of operators.

[0305] Step S5: Improve the rapid response and resource allocation mechanism

[0306] During the execution of ground support tasks, unexpected events and emergencies may occur at any time. In order to ensure that the system can respond quickly to these changes, Example 2 introduces an improved rapid response and resource allocation mechanism, which specifically includes the following sub-steps:

[0307] a. Hierarchical response system

[0308] The system divides ground support tasks into different levels (such as emergency, high priority, and regular) based on factors such as flight delay time, passenger demand, and weather conditions. Each level corresponds to a different resource allocation strategy to ensure that key tasks are handled first when resources are limited. By introducing a priority-based dynamic resource allocation algorithm, the system can automatically adjust resource allocation according to the task level and improve resource utilization efficiency.

[0309] b. Real-time monitoring and early warning platform

[0310] Using Internet of Things (IoT) technology and sensor networks, the system monitors the status of all ground support vehicles and service facilities in real time, including location, operating status, equipment health, etc. Through machine learning-based early warning models, the system can analyze monitoring data, identify potential risks (such as vehicle failures, equipment abnormalities, traffic congestion, etc.), and issue early warnings in a timely manner to prevent problems from occurring or mitigate their impact.

[0311] c. Standardized Operating Procedures (SOPs)

[0312] Detailed operational requirements, time limits and quality standards are formulated for each type of ground support mission to ensure that the mission is carried out efficiently according to the predetermined standards. Through the automated review system based on the knowledge graph, the system can monitor and review the execution of operating procedures in real time, ensure that all links strictly abide by SOPs, and improve the standardization and consistency of mission execution.

[0313] d. Multidisciplinary Teams (MDTs)

[0314] When dealing with major events or complex tasks, the system will form a multidisciplinary collaborative team (MDTs) consisting of multiple relevant departments. Through the task allocation system based on collaborative filtering, the system can optimize the collaborative efficiency of team members, ensure information sharing and collaborative work between departments, and thus solve emergencies quickly and efficiently.

[0315] e.Automation and intelligent auxiliary tools

[0316] The system is equipped with intelligent scheduling assistant software and wearable devices based on augmented reality (AR) technology, providing real-time guidance and task status display for ground crew. The intelligent scheduling assistant can automatically recommend the optimal resource allocation plan according to the current task and environmental conditions; the wearable device provides real-time task guidance and operation feedback for ground crew, improving work efficiency and accuracy.

[0317] Step S8: Application of self-learning algorithm

[0318] To improve the intelligence level and response ability of the system, Example 2 introduced a reinforcement learning algorithm based on Deep Deterministic Policy Gradient (DDPG) in Step S5. This algorithm can continuously optimize the scheduling strategy and path planning algorithm according to historical data and real-time data. Specifically, through continuous learning and adjustment, the system can adapt to different operating environments and emergencies, improving the decision-making ability and response speed of the system in complex and dynamic environments.

[0319] Step S9: Integrated decision support system

[0320] The system integrates an integrated decision support system based on Graph Neural Network (GNN), which can integrate information from modules such as intelligent prediction, resource optimization, path planning, and real-time adjustment. Using big data analysis and machine learning technologies, the integrated decision support system can deeply analyze and correlate the data of each module, improving the overall coordination of emergency response and the accuracy of decision-making. This system can generate comprehensive decision-making suggestions based on multi-source data to assist managers in making optimal decisions.

[0321] Step S10: Standard operating procedures and ground crew training

[0322] To ensure the efficient execution of ground support tasks and the operation proficiency of personnel, the system further includes the formulation of standard operating procedures (SOPs) and a ground crew training module. Through the operator training module based on virtual reality (VR) and augmented reality (AR) technologies, ground crew can conduct simulation training of system operations in an immersive training environment, improving their operation proficiency and emergency response ability for the automatic order dispatch system. In addition, through regular training and assessment, the system ensures that ground crew can proficiently master the latest operating procedures and technical requirements, maintaining a high level of work efficiency and service quality.

[0323] Working process

[0324] Task classification and resource allocation:

[0325] The system classifies ground support tasks according to urgency and priority based on real-time data analysis.

[0326] The priority-based dynamic resource allocation algorithm automatically adjusts resource allocation to ensure that high-priority tasks are processed first.

[0327] Real-time monitoring and risk warning: Through IoT technology and sensor networks, the system monitors the status of all ground support vehicles and equipment in real time. The warning model based on machine learning analyzes the monitoring data, identifies potential risks, and issues warnings in a timely manner.

[0328] Standardized operations and audits: The system formulates detailed operation requirements according to SOPs to ensure that tasks are executed according to standards. The automated audit system based on the knowledge graph monitors the operation process in real time to ensure the strict implementation of SOPs.

[0329] Collaborative team formation and task allocation: When major events occur, the system forms multi-disciplinary collaborative teams (MDTs). The task allocation system based on collaborative filtering optimizes team collaboration efficiency to ensure quick and efficient problem-solving.

[0330] Intelligent scheduling and real-time guidance: The intelligent scheduling assistant based on augmented reality (AR) technology provides real-time guidance for ground crew. Wearable devices display task status in real time, improving the work efficiency and accuracy of ground crew.

[0331] Self-learning and strategy optimization: The reinforcement learning algorithm based on DDPG continuously optimizes the scheduling strategy and path planning algorithm according to historical and real-time data. Through continuous learning, the system improves its decision-making ability and response speed in complex and dynamic environments.

[0332] Comprehensive decision support: The comprehensive decision support system integrates data from each module, conducts in-depth analysis through graph neural networks (GNNs), and generates comprehensive decision-making suggestions. Managers make optimal decisions based on system suggestions, improving the overall coordination and accuracy of emergency response.

[0333] Standardized operations and training: The system formulates detailed SOPs and conducts ground crew training through the operator training module of VR and AR technologies. Regular assessments ensure that ground crew are proficient in operating procedures and technical requirements, maintaining a high level of work efficiency and service quality.

[0334] The hierarchical response system and the priority-based dynamic resource allocation algorithm enable the system to prioritize critical tasks under limited resources, improving resource utilization efficiency.

[0335] The real-time monitoring and warning platform identifies potential risks through machine learning models, takes preventive measures in advance, and reduces the impact of emergencies on ground support tasks.

[0336] Standard Operating Procedures (SOPs) ensure that ground support tasks are carried out according to unified standards, and the implementation is monitored through an automated audit system, enhancing the standardization and consistency of task execution.

[0337] Multi-Disciplinary Teams (MDTs) optimize team collaboration efficiency through a collaborative filtering task assignment system, ensuring that problems can be solved quickly and efficiently in major events or complex tasks.

[0338] Intelligent scheduling assistants and wearable devices based on Augmented Reality (AR) technology provide real-time guidance and task status displays for ground crew, enhancing work efficiency and accuracy.

[0339] An integrated decision support system based on the Deep Deterministic Policy Gradient (DDPG) self-learning algorithm and Graph Neural Network (GNN) enables the system to continuously optimize scheduling strategies and path planning based on historical and real-time data, improving the system's intelligence level and response capabilities.

[0340] A security data transmission module based on blockchain ensures the integrity and security of data transmission. The predictive maintenance module predicts equipment failures in advance, reducing the impact of equipment failures on ground support tasks and enhancing the reliability of the system.

[0341] The user behavior analysis module optimizes the task assignment strategy, improving the overall efficiency of human-machine collaboration; the operator training module based on VR and AR technologies provides an immersive training environment, enhancing the operation proficiency and emergency response capabilities of ground crew, ensuring the efficient operation of the system.

[0342] By implementing this Embodiment 2, the system not only has the core function of automatic order assignment but also has been significantly enhanced in terms of security, equipment maintenance management, and personnel training. The improved rapid response and resource allocation mechanism, self-learning algorithm, and integrated decision support system enable the system to respond quickly in emergencies, allocate resources reasonably, and ensure the efficient execution of airport ground support tasks. In addition, the introduction of standard operating procedures and immersive training modules has enhanced the operation proficiency and emergency response capabilities of ground crew, further ensuring the reliability and service quality of the system. In summary, through multiple technological innovations, this Embodiment 2 has comprehensively improved the security, reliability, and intelligence level of the automatic order assignment method for civil aviation airport aircraft ground support, meeting the requirements of modern civil aviation airports for efficient, safe, and intelligent ground support services.

[0343] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present invention.

Claims

1. An automatic order assignment method based on aircraft ground support at civil aviation airports, characterized in that: It includes the following steps: S1. Use a time series prediction model to fuse multi-source data of historical flight data, weather forecasts, and air traffic control instructions, and utilize a dynamic causal relationship identification module based on a Bayesian network to identify the causal relationships between different data sources in real time to accurately predict flight arrival times; S2. Based on the prediction results of step S1, combined with real-time monitoring data, including flight dynamics, vehicle positions, and equipment status, use a context awareness module based on convolutional neural networks (CNNs) and long short-term memory networks (LSTMs) to dynamically adjust the priorities of tasks and perform dynamic task allocation; S3. Apply an improved multi-objective optimization algorithm to adaptively optimize the allocation of resources according to factors such as the urgency of tasks and the types of required equipment. This optimization algorithm combines the global search ability of genetic algorithms and the fast convergence characteristics of particle swarm optimization algorithms, and introduces a dynamic weight adjustment mechanism and a real-time feedback loop, with the ability of self-learning, and can continuously optimize the decision-making logic based on historical data and real-time feedback; S4. Use an improved path planning algorithm to plan the optimal path for each execution unit. The path planning algorithm is based on a model that combines multi-level A* search and deep reinforcement learning, can update in real time and adapt to the dynamic changes of the airport environment, and at the same time combines an energy management algorithm based on a prediction model to optimize energy consumption, ensuring that electric vehicles have enough energy to return to the base or reach the next charging point before completing the task; S5. Through a real-time monitoring and feedback mechanism, establish a feedback loop to feed the real-time monitoring data back to the learning module of the system. Use a self-learning algorithm based on deep deterministic policy gradients (DDPGs) and a comprehensive decision support system based on graph neural networks (GNNs) to dynamically adjust task allocation and path planning to ensure the overall progress of the system in case of emergencies, and through an improved rapid response and resource allocation mechanism, including a priority-based resource dynamic allocation algorithm, a machine learning-based early warning model, a knowledge graph-based automated audit system, a collaborative filtering-based task allocation system, and an augmented reality (AR)-technology-based intelligent dispatching assistant, quickly respond to emergencies and reasonably allocate resources; S6. Adopt a blockchain-based secure data transmission module to ensure the integrity and security of data during the multi-source data fusion and real-time data transmission processes, preventing data tampering and unauthorized access; S7. Integrate a predictive maintenance module to predict potential failures of ground support equipment based on real-time monitoring data and historical maintenance records, schedule maintenance tasks in advance, and reduce the impact of equipment failures on ground support tasks; S8. Through a user behavior analysis module, analyze the operation habits and task execution efficiency of ground crew members, optimize the task allocation strategy, and improve the overall efficiency of human-machine collaboration; S9. Adopt an operator training module based on virtual reality (VR) and augmented reality (AR) technologies to provide an immersive training environment for ground crew members and improve their proficiency in operating the automatic order dispatching system and emergency response capabilities; S10. Utilize big data analysis and machine learning techniques to deeply mine historical task execution data, optimize the task allocation model and resource configuration strategy, and continuously improve the overall performance and service quality of the system.

2. The automatic order assignment method based on aircraft ground support at civil aviation airports according to claim 1, wherein: The multi-source data fusion in step S1 includes time series data, meteorological image data, and air traffic control instruction text data, which are deeply integrated using a multi-modal data fusion algorithm based on deep learning, and the fusion ratio of different data sources is optimized using an adaptive weight allocation mechanism.

3. The automatic order assignment method based on aircraft ground support at civil aviation airports according to claim 1, wherein: The situation awareness module in step S2 uses a multi-modal data fusion algorithm based on convolutional neural network (CNN) and long short-term memory network (LSTM), and uses Internet of Things (IoT) technology and sensor networks to obtain and integrate relevant data in real time to accurately perceive the current operating status of the airport and changes in the external environment.

4. The automatic order assignment method based on aircraft ground support at civil aviation airports according to claim 1, wherein: The improved multi-objective optimization algorithm in step S3 combines the global search ability of the genetic algorithm with the fast convergence characteristics of the particle swarm optimization algorithm, and introduces a dynamic weight adjustment mechanism, which improves the computational efficiency and solution accuracy of the optimization and adapts to the requirements of complex and changeable ground support tasks.

5. The automatic order assignment method based on aircraft ground support at civil aviation airports according to claim 1, characterized in that: The improved path planning algorithm in step S4 combines the heuristic search of the multi-level A* search algorithm with the policy optimization of the deep reinforcement learning method, can plan and adjust the path in real time in a complex airport environment, dynamically avoid static and dynamic obstacles, and improve the efficiency and accuracy of path planning through hierarchical path segment optimization.

6. The automatic order assignment method based on aircraft ground support at civil aviation airports according to claim 1, wherein: The energy consumption optimization in step S4 further includes real-time monitoring of the battery status of electric vehicles, and using an energy management algorithm based on a prediction model to dynamically adjust the path planning to minimize energy consumption and charging times, and ensure that the vehicle has enough energy to return to the base or reach the next charging point before completing the task.

7. An automatic order dispatching method based on aircraft ground support at civil aviation airports according to claim 1, characterized in that: The improved rapid response and resource allocation mechanism in step S5 includes: a. A hierarchical response system that classifies ground support tasks into different levels according to factors such as flight delay time, passenger demand, and weather conditions, and introduces a priority-based dynamic resource allocation algorithm; b. A real-time monitoring and early warning platform that uses IoT technology and sensor networks to monitor the status of all ground support vehicles and service facilities in real time, and identifies potential risks through a machine learning-based early warning model; c. Standard operating procedures (SOPs) that formulate detailed operation requirements, time limits, and quality standards for each type of ground support task, and ensure the strict implementation of SOPs through a knowledge graph-based automated audit system; d. Multi-disciplinary collaboration teams (MDTs) composed of multiple relevant departments, responsible for handling major events or complex tasks, and optimizing the team collaboration efficiency through a collaborative filtering-based task allocation system; e. Automated and intelligent auxiliary tools, including intelligent scheduling assistant software based on augmented reality (AR) technology and wearable devices, to provide real-time guidance and task status display for ground crew.

8. The automatic order assignment method based on aircraft ground support at civil aviation airports according to claim 1, wherein: The self-learning algorithm in step S5 is selected from a reinforcement learning algorithm based on Deep Deterministic Policy Gradient (DDPG), which can continuously optimize the scheduling strategy and path planning algorithm according to historical and real-time data, improving the intelligence level and response ability of the system.

9. The automatic order assignment method based on aircraft ground support at civil aviation airports according to claim 1, characterized in that: The comprehensive decision support system integrates the information of modules for intelligent prediction, resource optimization, path planning, and real-time adjustment, adopts a unified decision-making framework based on Graph Neural Network (GNN), and utilizes big data analysis and machine learning technologies to enhance the overall coordination of emergency response and the accuracy of decision-making.

10. The automatic order assignment method based on aircraft ground support at civil aviation airports according to claim 1, wherein: The automatic dispatch system further includes a module for formulating Standard Operating Procedures (SOPs) and training ground crew, ensuring that various ground support tasks are efficiently executed according to predetermined standards. Additionally, through an operator training module based on Virtual Reality (VR) and Augmented Reality (AR) technologies, an immersive training environment is provided for ground crew, improving their proficiency in operating the automatic dispatch system and their emergency response capabilities.

Images