Multi-flight scene avionics service dynamic scheduling method and system
By constructing a multi-flight scenario library and a reinforcement learning reward function, the avionics system resources are dynamically scheduled, solving the resource scheduling problem of the avionics system in multiple scenarios and improving the adaptability and reliability of the aircraft.
Patent Information
- Application Number
- CN202511841673.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-09
- Publication Date
- 2026-03-31
AI Technical Summary
Existing avionics system resource scheduling technologies have failed to effectively cope with dynamic changes in multiple flight scenarios, resulting in insufficient or wasted resources. Furthermore, the lack of real-time demand classification and closed-loop optimization affects the safety, economy, and comfort of aircraft.
A multi-flight scenario library is constructed, and avionics functions are divided into strong real-time, medium real-time, and weak real-time sub-services. Real-time scheduling is achieved through reinforcement learning reward functions, combined with a distributed registry center and closed-loop optimization mechanism to dynamically match resource requirements.
It improves the adaptability and reliability of avionics systems in multiple scenarios, reduces the risk of critical service failures, optimizes resource utilization, and balances safety and economy.
Smart Images

Figure CN121764616A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of integrated avionics technology, specifically relating to a dynamic scheduling method and system for avionics services in multiple flight scenarios based on reinforcement learning reward functions. Background Technology
[0002] As civil aircraft become larger and more intelligent, avionics systems integrate increasingly complex functions such as flight control, navigation, communication, and cabin management. Their core requirements have evolved from single-function implementation to efficient multi-task collaborative operation. The stable operation of avionics systems highly depends on the rational allocation of resources, and resource requirements vary significantly across different flight scenarios. For example, during takeoff, priority must be given to ensuring low latency for high-real-time services such as flight control command parsing; during cruise, a balance must be struck between resource utilization and the experience of low-real-time services such as cabin entertainment; and under extreme weather conditions, the system's fault tolerance must be enhanced. Therefore, resource scheduling strategies directly impact the safety, economy, and comfort of aircraft.
[0003] However, existing avionics resource scheduling technologies have significant limitations: traditional scheduling is mostly based on static allocation of resources according to preset rules, without considering the dynamic changes in flight phases and flight conditions, which can easily lead to insufficient resources in critical phases or waste of resources in non-critical phases; avionics functions are not clearly classified according to real-time requirements, and strong real-time services and weak real-time services share the resource pool, which may lead to weak real-time services preempting strong real-time services and causing critical service response delays to exceed limits; at the same time, there is a lack of a closed-loop optimization mechanism based on real-time performance indicators. When there are abnormal resource utilization, critical service delays exceeding limits, or fault recovery timeouts, the scheduling strategy cannot be adaptively adjusted, requiring manual intervention, which is not flexible enough. Furthermore, it does not design differentiated resource mapping rules for complex combinations of "flight phases + flight conditions", making it difficult to balance safety and resource efficiency.
[0004] To address the dynamic resource scheduling needs of avionics systems in multiple scenarios, there is an urgent need for a technical solution that can sense flight phases and conditions, accurately match sub-service resource requirements, and optimize scheduling strategies through real-time feedback, so as to improve the adaptability, reliability, and resource utilization efficiency of avionics systems. Summary of the Invention
[0005] The purpose of this invention is to provide a dynamic scheduling method and system for avionics services in multiple flight scenarios, which solves the problem of difficult dynamic resource scheduling of avionics systems in multiple scenarios. Based on flight scenarios, the invention provides a technical solution that accurately matches the resource requirements of sub-services and optimizes scheduling strategies through real-time feedback, thereby improving the adaptability, reliability and resource utilization efficiency of avionics systems.
[0006] The technical solution of this invention is implemented as follows: In a first aspect, the present invention provides a method for dynamic scheduling of avionics services in multiple flight scenarios, which includes the following steps: Step 1: Construct a flight scenario library, which includes the first flight stage, the second flight stage, the third flight stage, the fourth flight stage, as well as Class I flight conditions, Class II flight conditions, and Class III flight conditions, and define the judgment conditions for each scenario; Step 2: Based on the real-time requirements of avionics functions, complex avionics functions are divided into three sub-services: Category 1, Category 2, and Category 3. Each sub-service is bound to a unique resource requirement tag. Step 3: Generate standardized service instances from the split Class I, Class II, and Class III sub-services, and register them synchronously with the service registry center; Step 4: For multiple flight scenarios in the flight scenario library, match sub-services and define scenario-service-resource quantization mapping rules; Step 5: Based on historical flight data, select a sample set of flight scenario databases that includes the first flight phase, the second flight phase, the third flight phase, the fourth flight phase, and Class I, Class II, and Class III flight conditions; Step 6: Using the first feature, second feature, and third feature as the constraint targets, design a reinforcement learning reward function under multiple flight scenario conditions; Step 7: Real-time acquisition of flight phase judgment signals, flight condition judgment signals, first feature, second feature, and third feature operational performance indicators; and dynamic scheduling of avionics services through a reward function. Step 8: Determine if the performance indicators are normal. If not, adjust the corresponding weights in the reward function and return to Step 6; if yes, end the process.
[0007] As a further aspect of the present invention: the first flight phase is a taxiing phase, the second flight phase is a takeoff phase, the third flight phase is a cruise phase, and the fourth flight phase is a landing phase; The first category of flight conditions refers to extremely severe weather conditions, the second category of flight conditions refers to normal complex weather conditions, and the third category of flight conditions refers to clear-sky standard weather conditions.
[0008] As a further aspect of the present invention: the sub-service is a strong real-time sub-service with a response latency threshold in milliseconds; The second type of sub-service is a medium real-time sub-service with a response latency threshold in the hundreds of milliseconds. The three types of sub-services are weak real-time sub-services, with a response latency threshold in the second range.
[0009] As a further aspect of the present invention: In step three, the service registry adopts a distributed registry based on ETCD to store the metadata of the service instance, including service type, resource requirements, and the general computing module in which it resides.
[0010] As a further aspect of the present invention: in step six, the first feature is resource utilization, the second feature is critical service response delay, and the third feature is fault recovery time.
[0011] As a further aspect of the present invention: in step six, the expression of the reward function is:
[0012] in, U represents the resource utilization score, D represents the penalty for exceeding the service response latency limit, and F represents the penalty for exceeding the fault recovery time limit.
[0013] As a further aspect of the present invention: In step eight, the method for adjusting the corresponding weights in the reward function is as follows: if the resource utilization rate is low, then the weights are increased. Weighting will be adjusted upwards if the response latency of critical services exceeds the limit. The weighting is adjusted upwards if the fault recovery time exceeds the standard. The weights are determined; an adaptive scheduling strategy is implemented based on the feedback of the reward function.
[0014] As a further aspect of the present invention: the judgment signals of flight phase and flight conditions are acquired in real time through the avionics bus interface.
[0015] As a further aspect of the present invention: the judgment conditions for each scenario: Category 1 conditions: Based on data from onboard weather radar and ground control towers, the following conditions must be met: crosswind ≥ 20 m / s, visibility ≤ 500 meters, or thunderstorm echoes detected. Category II conditions: Meeting the requirements of "crosswind 10~20m / s", "visibility 500~1000 meters", or "light rain / light fog"; Three conditions must be met: "crosswind ≤ 10m / s", "visibility ≥ 1000 meters", and "no precipitation / low clouds".
[0016] Secondly, the present invention provides a dynamic scheduling system for avionics services in multiple flight scenarios, comprising: The scenario building module constructs a flight scenario library that includes four flight phases (taxiing, takeoff, cruise, and landing) and three types of flight conditions (extremely severe weather, normal complex weather, and clear sky standard weather). Based on real-time requirements, the avionics functions are divided into three types of sub-services: strong real-time, medium real-time, and weak real-time, and registered to the distributed service center. The mapping module constructs a flight scenario library that includes four flight phases (taxiing, takeoff, cruise, and landing) and three types of flight conditions (extremely severe weather, normal complex weather, and clear sky standard weather). Based on real-time requirements, it breaks down avionics functions into three types of sub-services: strong real-time, medium real-time, and weak real-time, and registers them to the distributed service center. The scheduling module collects operational metrics in real time and dynamically adjusts the weight of the reward function to form a closed-loop scheduling mechanism.
[0017] The beneficial effects of this invention are as follows: 1. This invention designs and develops a dynamic scheduling method for avionics services in multiple flight scenarios based on reinforcement learning reward functions. By constructing a multi-flight scenario library and combining sub-service real-time classification and differentiated resource mapping rules, the avionics services are dynamically matched to the core requirements of the scenario, significantly enhancing the adaptability of the avionics system in the entire flight process.
[0018] 2. This invention effectively reduces the risk of critical service failure and improves the reliability of avionics systems by defining a reinforcement learning reward function and core constraint objectives, and dynamically adjusting weights by combining a closed-loop optimization mechanism.
[0019] 3. This invention dynamically adjusts the weights of the reward function to guide resources towards less real-time sub-services in non-critical scenarios, avoiding resource idleness; and reduces resource waste through precise allocation in critical scenarios. Compared with traditional static scheduling, it can reduce energy consumption and balance security and economy.
[0020] 4. This invention achieves dynamic adaptation of avionics resources through multi-flight scenario perception and intelligent reward mechanism, effectively improving resource utilization, ensuring the real-time performance of critical services, and enhancing the adaptability and reliability of avionics systems in complex scenarios. It is applicable to aircraft avionics systems.
[0021] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0022] Figure 1 This is a flowchart illustrating a dynamic scheduling method for avionics services in multiple flight scenarios based on a reinforcement learning reward function, according to the present invention. Detailed Implementation
[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions in the embodiments of this invention will be described in more detail below with reference to the accompanying drawings.
[0024] In the accompanying drawings, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The described embodiments are some, but not all, of the embodiments of the present invention.
[0025] The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.
[0026] The following is in conjunction with the appendix Figure 1 The embodiments of the present invention will be described in detail below.
[0027] Example 1 This invention discloses a dynamic scheduling method for avionics services in multiple flight scenarios based on a reinforcement learning reward function, which includes the following steps: Step 1: Construct a flight scenario library, which includes the first flight stage, the second flight stage, the third flight stage, the fourth flight stage, as well as Class I flight conditions, Class II flight conditions, and Class III flight conditions, and define the judgment conditions for each scenario.
[0028] Step 2: Based on the real-time requirements of avionics functions, complex avionics functions are divided into three sub-services: Category 1, Category 2, and Category 3. Each sub-service is bound to a unique resource requirement tag.
[0029] Step 3: Generate standardized service instances from the split Class I, Class II, and Class III sub-services, and register them synchronously with the service registry center.
[0030] Step 4: For multiple flight scenarios in the flight scenario library, match sub-services and define scenario-service-resource quantization mapping rules.
[0031] Step 5: Based on historical flight data, select a sample set of flight scenarios that includes the first flight phase, the second flight phase, the third flight phase, the fourth flight phase, and the first, second, and third flight conditions.
[0032] Step 6: Using the first feature, second feature, and third feature as the constraint targets, design a reinforcement learning reward function under multiple flight scenarios.
[0033] Step 7: Real-time acquisition of flight phase judgment signals, flight condition judgment signals, first feature, second feature, and third feature operational performance indicators, and dynamic scheduling of avionics services through a reward function.
[0034] Step 8: Determine if the performance indicators are normal. If not, adjust the corresponding weights in the reward function and return to Step 6; if yes, end the process.
[0035] Preferably, the first flight phase is the taxiing phase, the second flight phase is the takeoff phase, the third flight phase is the cruise phase, and the fourth flight phase is the landing phase; the first type of flight condition is extreme adverse weather conditions, the second type of flight condition is normal complex weather conditions, and the third type of flight condition is clear-sky standard weather conditions.
[0036] Preferably, the first type of sub-service is a strong real-time sub-service with a response latency threshold in milliseconds; the second type of sub-service is a medium real-time sub-service with a response latency threshold in hundreds of milliseconds; and the third type of sub-service is a weak real-time sub-service with a response latency threshold in seconds.
[0037] Preferably, in step three, the service registry is an ETCD-based distributed registry used to store the metadata of the service instance, including service type, resource requirements, and the general computing module in which it resides.
[0038] Preferably, in step six, the first feature is resource utilization, the second feature is critical service response delay, and the third feature is fault recovery time.
[0039] Preferably, in step six, the expression for the reward function is:
[0040] in, U represents the resource utilization score, D represents the penalty for exceeding the service response latency limit, and F represents the penalty for exceeding the fault recovery time limit.
[0041] Preferably, in step eight, the method for adjusting the corresponding weights in the reward function is as follows: if resource utilization is low, then the weights are increased. Weighting will be adjusted upwards if the response latency of critical services exceeds the limit. The weighting is adjusted upwards if the fault recovery time exceeds the standard. The weights are determined; an adaptive scheduling strategy is implemented based on the feedback of the reward function.
[0042] Preferably, it supports real-time acquisition of flight phase and flight condition judgment signals via the avionics bus interface.
[0043] As civil aircraft become larger and more intelligent, avionics systems integrate increasingly complex functions such as flight control, navigation, communication, and cabin management. Their core requirements have evolved from single-function implementation to efficient multi-task collaborative operation. The stable operation of avionics systems highly depends on the rational allocation of resources, and resource requirements vary significantly across different flight scenarios. For example, during takeoff, priority must be given to ensuring low latency for high-real-time services such as flight control command parsing; during cruise, a balance must be struck between resource utilization and the experience of low-real-time services such as cabin entertainment; and under extreme weather conditions, the system's fault tolerance must be enhanced. Therefore, resource scheduling strategies directly impact the safety, economy, and comfort of aircraft. However, existing avionics resource scheduling technologies have significant limitations: traditional scheduling is mostly based on static resource allocation according to preset rules, without considering the dynamic changes in flight phases and conditions, which can easily lead to insufficient resources in critical phases or waste of resources in non-critical phases; avionics functions are not clearly classified according to real-time requirements, and strong real-time services and weak real-time services share the same resource pool, which may result in weak real-time services preempting resources of strong real-time services and causing excessive response delays in critical services; at the same time, there is a lack of a closed-loop optimization mechanism based on real-time performance indicators. When abnormal resource utilization, excessive delays in critical services, or timeouts in fault recovery occur, the scheduling strategy cannot be adaptively adjusted, requiring manual intervention, which is not flexible enough, and there is no design for differentiated resource mapping rules for complex combinations of "flight phase + flight conditions", making it difficult to balance safety and resource efficiency. To address the dynamic resource scheduling needs of avionics systems in multiple scenarios, there is an urgent need for a technical solution that can sense flight phases and conditions, accurately match sub-service resource requirements, and optimize scheduling strategies through real-time feedback, in order to improve the adaptability, reliability, and resource utilization efficiency of avionics systems.
[0044] Example 2 This invention discloses a dynamic scheduling method for avionics services in multiple flight scenarios based on a reinforcement learning reward function, which includes the following steps: Step 1: Construct a flight scenario library, which includes the first flight stage, the second flight stage, the third flight stage, the fourth flight stage, as well as Class I flight conditions, Class II flight conditions, and Class III flight conditions, and define the judgment conditions for each scenario.
[0045] Flight phase: Phase 1 of flight: from engine start-up (engine speed ≥ 50%) to entering the runway takeoff line; Second phase of flight: After entering the runway takeoff line and taking off, ascend to an altitude of 1000 meters; Third flight phase: Ascend to a cruising altitude of 8,000 meters and maintain level flight until descent begins; Fourth flight phase: After the aircraft begins its descent and touchdown, it taxis to the parking position and the engine speed is reduced to idle.
[0046] Flight conditions: Category 1 conditions: Based on data from onboard weather radar and ground control towers, the following conditions must be met: crosswind ≥ 20 m / s, visibility ≤ 500 meters, or thunderstorm echoes detected. Category II conditions: Meeting the requirements of "crosswind 10~20m / s", "visibility 500~1000 meters", or "light rain / light fog"; Three conditions must be met: "crosswind ≤ 10m / s", "visibility ≥ 1000 meters", and "no precipitation / low clouds".
[0047] Step 2: Based on the real-time requirements of avionics functions, complex avionics functions are divided into three sub-services: Category 1, Category 2, and Category 3. Each sub-service is bound to a unique resource requirement tag.
[0048] One type of sub-service: requires strong real-time response, with millisecond-level response time; Category II sub-services: Medium real-time, requiring millisecond-level response time; Three types of sub-services: weak real-time, requiring millisecond-level response; Step 3: Generate standardized service instances from the split Class I, Class II, and Class III sub-services, and register them synchronously with the service registry center.
[0049] The distributed service registry based on ETCD registers the metadata (service type, resource requirements, and the general computing module in which they reside) of all instances.
[0050] APP1 is a first-class sub-service, APP2 is a second-class sub-service, and APP3 is a third-class sub-service.
[0051] <APP1ComputeTime="75"PartitionId="1"Period="100"RequiredMemorySize="100000"Priority="1"Hardware="CPM"> <APP2ComputeTime="40"PartitionId="2"Period="100"RequiredMemorySize="100000"Priority="2"Hardware="CPM"> <APP3ComputeTime="35"PartitionId="3"Period="100"RequiredMemorySize="50000"Priority="3"Hardware="CPM"> Step 4: For multiple flight scenarios in the flight scenario library, match sub-services and define scenario-service-resource quantization mapping rules.
[0052] Scenario 1: First flight phase + Category I flight conditions Matching sub-services: Category 1 sub-services, Category 2 sub-services, and Category 3 sub-services; Sub-service priority: Category 1 > Category 2 > Category 3. When there is a resource conflict, Category 2 can preempt Category 3, and Category 1 can preempt Category 2 and Category 3.
[0053] Step 5: Based on historical flight data, select a sample set of flight scenarios that includes the first flight phase, the second flight phase, the third flight phase, the fourth flight phase, and the first, second, and third flight conditions.
[0054] Step 6: Using the first feature, second feature, and third feature as the constraint targets, design a reinforcement learning reward function under multiple flight scenarios.
[0055] Design a reward function with resource utilization, critical service response latency, and fault recovery time as constraints.
[0056] U represents the resource utilization score, ranging from 0 to 1. When the resource utilization rate is 60% to 80%, U=1, and it decreases linearly when it is below 60% or above 80%. D is the penalty for exceeding the service response latency limit, ranging from 0 to 1. When the latency is less than or equal to the threshold, D = 0. Exceeding the limit is calculated linearly according to the "exceedance ratio". F is the fault recovery timeout penalty, ranging from 0 to 1. When the recovery time is less than or equal to the threshold, F = 0. The timeout is calculated linearly according to the "timeout ratio". The weights are dynamically adjusted according to the scenario, and the thresholds are based on the average performance index of the constrained targets in the sample set, with a reserved safety margin.
[0057] Step 7: Real-time acquisition of flight phase judgment signals, flight condition judgment signals, first feature, second feature, and third feature operational performance indicators, and dynamic scheduling of avionics services through a reward function.
[0058] Scenario 1 Service Scheduling: Real-time signal acquisition: Triggering "first flight phase + Class I conditions" (500 meters after takeoff, crosswind 22 m / s); Initial weights: =0.1, =0.6, =0.3; Step 8: Determine if the performance indicators are normal. If not, adjust the corresponding weights in the reward function and return to Step 6; if yes, end the process.
[0059] Initial scheduling strategy: APP1 occupies CPM1 computing resources, while APP2 and APP3 share CPM2 computing resources.
[0060] Scenario 1 optimization: Real-time performance metrics collected: APP1 service response latency is 30ms, threshold is 20ms, mapping D=0.5; resource utilization is 75%, calculation U=1; no faults, F=0; Calculate the reward function
[0061] Adjusting weights: =0.05, =0.7, =0.25; After optimization: APP3's resource usage on CPM2 is released; APP1 uses CPM1 and CPM2 computing resources, and APP2 uses CPM2 computing resources; APP1's service response latency is 18ms, and mapping D=0; Calculate the reward function ; Return to normal, complete the dynamic scheduling of avionics services in this scenario, and continuously monitor operational performance indicators.
[0062] Example 3 This invention discloses a dynamic scheduling system for avionics services in multiple flight scenarios, comprising: The scenario building module constructs a flight scenario library that includes four flight phases (taxiing, takeoff, cruise, and landing) and three types of flight conditions (extremely severe weather, normal complex weather, and clear sky standard weather). Based on real-time requirements, the avionics functions are divided into three types of sub-services: strong real-time, medium real-time, and weak real-time, and registered to the distributed service center. The mapping module constructs a flight scenario library that includes four flight phases (taxiing, takeoff, cruise, and landing) and three types of flight conditions (extremely severe weather, normal complex weather, and clear sky standard weather). Based on real-time requirements, it breaks down avionics functions into three types of sub-services: strong real-time, medium real-time, and weak real-time, and registers them to the distributed service center. The scheduling module collects operational metrics in real time and dynamically adjusts the weight of the reward function to form a closed-loop scheduling mechanism.
[0063] By using multi-flight scenario perception and intelligent reward mechanisms, dynamic adaptation of avionics resources can be achieved, effectively improving resource utilization, ensuring the real-time performance of critical services, and enhancing the adaptability and reliability of avionics systems in complex scenarios. This technology is applicable to aircraft avionics systems.
[0064] Thus, the objective of this invention has been achieved.
[0065] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for dynamic scheduling of multi-flight scenario avionics services, characterized in that, The method comprises the following steps: Step 1: Construct a flight scene library, which contains a first flight phase, a second flight phase, a third flight phase, a fourth flight phase, a first flight condition, a second flight condition, and a third flight condition, and defines the judgment conditions of each scene; Step 2: According to the real-time requirement of the avionics function, the complex avionics function is divided into a first sub-service, a second sub-service, and a third sub-service, and each type of sub-service is bound to a unique resource demand tag; Step 3: The first sub-service, the second sub-service, and the third sub-service after splitting are generated into standardized service instances and are simultaneously registered to a service registration center; Step 4: For multiple flight scenes in the flight scene library, the sub-services are matched, and scene-service-resource quantization mapping rules are defined; Step 5: Based on historical flight data, a flight scene library sample set containing the first flight phase, the second flight phase, the third flight phase, the fourth flight phase, the first flight condition, the second flight condition, and the third flight condition is screened; Step 6: With the first feature, the second feature, and the third feature as constraint targets, a reinforcement learning reward function under multiple flight scene conditions is designed; Step 7: Real-time collection of flight phase judgment signals, flight condition judgment signals, first feature, second feature, and third feature performance indicators is performed, and through the reward function, dynamic scheduling of avionics services is realized; Step 8: It is judged whether the performance indicators are normal, if not, the corresponding weight in the reward function is adjusted, and the process returns to step 6; if yes, the process ends.
2. The method of claim 1, wherein, The first flight phase is a taxiing phase, the second flight phase is a takeoff phase, the third flight phase is a cruising phase, and the fourth flight phase is a landing phase; The first flight condition is an extreme severe weather condition, the second flight condition is a regular complex weather condition, and the third flight condition is a clear sky standard weather condition.
3. The method of claim 1, wherein, The first sub-service is a strong real-time sub-service, and the response delay threshold is millisecond level; The second sub-service is a medium real-time sub-service, and the response delay threshold is hundred-millisecond level; The third sub-service is a weak real-time sub-service, and the response delay threshold is second level.
4. The method of claim 1, wherein, In step 3, the service registration center adopts an ETCD-based distributed registration center for storing metadata of service instances, including service type, resource demand, and general computing module.
5. The method of claim 1, wherein, In step 6, the first feature is resource utilization, the second feature is key service response delay, and the third feature is fault recovery time.
6. The method of claim 1, wherein, In step 6, the expression of the reward function is: wherein, U is a resource utilization score, D is a service response delay penalty, and F is a failure recovery time penalty.
7. The method of claim 6, wherein, In step eight, the corresponding weights in the reward function are adjusted as follows: if resource utilization is low, the weights are increased. Weighting will be adjusted upwards if the response latency of critical services exceeds the limit. The weighting is adjusted upwards if the fault recovery time exceeds the standard. The weights are determined; an adaptive scheduling strategy is implemented based on the feedback of the reward function.
8. The method of claim 1, wherein, The judgment signals of the flight phase and the flight condition are obtained in real time through an avionics bus interface.
9. The method of claim 2, wherein, The judgment conditions of each scene are: The first condition: through the on-board weather radar and the ground tower data, "crosswind ≥ 20 m / s" or "visibility ≤ 500 meters" or "detecting thunderstorm echo" is satisfied; The second condition: "crosswind 10~20 m / s" or "visibility 500~1000 meters" or "light rain / light fog" is satisfied; The third condition: "crosswind ≤ 10 m / s" and "visibility ≥ 1000 meters" and "no precipitation / low cloud" are satisfied. 10.A multi-flight scenario avionics service dynamic scheduling system, characterized in that, The method comprises the following steps: The scene building module builds a flight scene library containing four flight stages of taxiing, taking off, cruising and landing and three flight conditions of extremely severe weather, regular complex weather and clear standard weather, and splits the avionics function into three types of sub-services of strong real-time, medium real-time and weak real-time according to real-time requirements and registers to the distributed service center. The mapping module builds a flight scene library containing four flight stages of taxiing, taking off, cruising and landing and three flight conditions of extremely severe weather, regular complex weather and clear standard weather, and splits the avionics function into three types of sub-services of strong real-time, medium real-time and weak real-time according to real-time requirements and registers to the distributed service center. The scheduling module collects running indexes in real time and dynamically adjusts the weight of the reward function to form a closed-loop scheduling.
