Online convex optimization method and system for HTN task planning of unmanned aerial vehicle

Abstract

The invention provides an online convex optimization method and system for unmanned aerial vehicle HTN task planning, relates to the field of artificial intelligence and intelligent planning, and adopts an online convex optimization (OCO) technology to adjust the time cost and resource consumption of task execution in real time according to the task planning requirements of an unmanned aerial vehicle in an uncertain and dynamic environment. According to the method, a task model is established through PDDL, and a task decomposition structure and target constraints are defined; decomposing the complex task into a plurality of executable actions (actions) by utilizing an HTN (Hypertext Transport Network) planning algorithm; a convex set is constructed based on loss in action execution, a cost index is defined, and action consumption is optimized online; and dynamically updating indexes according to actual feedback, adjusting subsequent action strategies, iteratively reducing consumption, and finally generating an optimized real-time plan (plan). Different from traditional static planning or off-line optimization, the method remarkably improves the action execution efficiency of the unmanned aerial vehicle, reduces resource and time waste, and can flexibly adapt to actual changes.

Description

Technical Field

[0001] The present invention belongs to the fields of artificial intelligence and intelligent planning, and particularly relates to an online convex optimization method and system for UAV HTN mission planning. Background Art

[0002] The rapid development of the low-altitude economy is closely related to the introduction of autonomous aircraft mission planning. The introduction of autonomous aircraft mission planning aims to improve the operation safety and efficiency in multiple fields such as environmental monitoring, logistics transportation, agricultural spraying, and disaster relief through intelligent and automated means. However, the problem of inability to plan tasks online may directly lead to the failure of autonomous aircraft mission planning and waste of resources and time. Therefore, it is particularly important to build an efficient and dynamic autonomous aircraft mission planning method, which is crucial for ensuring the efficiency and accuracy of task completion.

[0003] Traditional mission planning problems often adopt methods such as classical planning algorithms, HTN planning algorithms, and heuristic-based methods. However, traditional planning algorithms have poor dynamic performance, lack real-time evaluation of actual resources and time, and are limited in dealing with uncertain environments, making it difficult to ensure the efficiency and accuracy of task execution.

[0004] During the mission planning process of autonomous aircraft, the following challenges are mainly faced:

[0005] (1) During the mission planning process of autonomous aircraft, since the planning algorithm is often a static solution process, there are significant challenges in online mission planning in the actual flight environment. The HTN planning algorithm may perform best when dealing with specific types of tasks in a static environment, but its results may differ from actual task execution.

[0006] (2) During the mission planning process of autonomous aircraft, the results of mission planning do not consider the uncertainties of actual flight, which poses a significant challenge. When an autonomous aircraft executes a mission, it may cause varying degrees of waste of resources and time due to the actual mission, which may not only lead to an increase in mission costs and a decrease in efficiency, but also directly result in mission failure. Summary of the Invention

[0007] The invention purpose of the present invention is: aiming at all or part of the above-mentioned problems, to provide an online convex optimization method and system for autonomous aircraft HTN mission planning, so as to more efficiently and accurately improve the resource utilization rate and time management ability of autonomous aircraft.

[0008] The technical solution adopted by the present invention is as follows:

[0009] An online convex optimization method for UAV HTN mission planning, which includes:

[0010] S1. Based on the mission objectives and constraints faced by the autonomous aircraft, use PDDL to establish a mission model, clarify the mission objectives, resource requirements and constraints, and ensure that the model can accurately describe the mission execution environment and requirements;

[0011] S2. Utilize the HTN algorithm to decompose the complex mission into multiple actions and generate a preliminary plan, thereby simplifying the complexity of mission execution;

[0012] S3. According to the time and resources required during the execution of the autonomous aircraft's actions, construct a convex set to represent the time and resource constraints of the actions, and design a loss cost metric to quantify the total resource consumption and time consumption of the actions;

[0013] S4. Optimize the time and resource allocation of the actions through the GD optimization algorithm to minimize the loss cost metric, thereby reducing the total time and resource consumption of action execution and improving the action execution efficiency;

[0014] S5. According to the feedback data during the actual action execution of the autonomous aircraft, dynamically update the loss cost metric and adjust the execution strategy of subsequent actions to adapt to environmental changes and mission requirements, and ensure the flexibility and real-time nature of mission execution;

[0015] S6. Generate the final optimized plan, and through a closed-loop optimization mechanism, achieve real-time update and dynamic adjustment of action execution, and ensure that the actions can be efficiently completed and adapt to the changes in the dynamic environment.

[0016] Further, S1 includes:

[0017] S1.1. According to the mission to be executed by the autonomous aircraft, define the specific mission objectives of the autonomous aircraft, and formulate constraints such as time limits and resource limits;

[0018] S1.2. Use the planning domain language to establish the decomposition structure of the mission, define complex missions and actions and their dependencies. Describe the resources required for each complex mission and action, and set the estimated execution time for them.

[0019] Further, S2 includes:

[0020] S2.1. Utilize the HTN algorithm to decompose the complex mission into multiple levels of actions;

[0021] S2.2. Select the task decomposition path according to the current task requirements to ensure that the plan meets the current task objectives and constraints, and generate a plan consisting entirely of actions.

[0022] Further, S3 includes:

[0023] S3.1. Divide each action in the plan into a resource set and a time set according to its resource consumption and time consumption. Construct a convex set with the data in the resource set and the time set, and define the boundary conditions of the convex set to ensure that the action execution is within the feasible region.

[0024] The convex set consists of the set of resource consumption T = {t A , t B ,....., t N} and the set of time consumption R = {r A , r B ,....., r N} of all possible actions, which is represented as a range in a multi-dimensional space, and its expression is as follows:

[0025] X = {(t A , t B ,....., t N , r a , r b ,....., r n ) | t min <= t A , t B ,....., t N <= t max , r min <= r a , r b ,....., r n <= r max}

[0026] This set is a convex set if and only if for any two points x1, x2 ∈ S in the set, the line segment connecting these two points belongs to S, that is, it satisfies:

[0027]

[0028] S3.2. Use the action time and resource constraints in the convex set to construct a loss cost index for the action to measure the total resource consumption and time consumption of the action. Take the loss cost index as a quantitative goal, and evaluate the execution efficiency of the autonomous aircraft in different action plans according to this loss cost index.

[0029] The loss cost metric is a convex function that measures the cost of a decision, the loss cost

[0030] The metric formula is as follows:

[0031] L(t,r) = ∑c i *t i +∑a i *r i +λ∑(t j -t i ) 2

[0032] Where c i represents the importance coefficient of execution time, a i represents the importance coefficient of resource consumption, (t j -t i ) 2 represents the time-dependent constraint between actions, λ is the penalty coefficient for the time difference between actions, which controls the continuity of action execution. The first two weighted sums represent the consumption of resources and time. The last term represents the time interval between action executions, ensuring the continuity of time allocation.

[0033] Furthermore, S4 includes:

[0034] S4.1. Set the initial parameters in the GD optimization algorithm, control the amplitude of parameter changes in each update according to the adjustment factor η, and set the condition for the algorithm to stop optimizing through the number of iterations Max_iter. Among them, η determines the step size of the change rate decrease, which is used to control the amplitude of each parameter update. By determining the appropriate value of η, the optimization path will quickly and stably reach the optimum, and Max_iter limits the number of optimization iterations.

[0035] S4.2. Based on the current plan and the loss cost metric, calculate the partial derivatives of the loss cost metric with respect to action time and resources to obtain the change rate information. The change rate reflects the sensitivity of the current action time to the support configuration. The change rate calculation formula is:

[0036]

[0037] Calculate the change rate of each action time decision, which represents the impact of the loss cost metric of the original action on the time change rate.

[0038]

[0039] According to the currently preset resource decision, calculate its change rate, which describes the impact of the loss cost metric of this action on the resource change rate.

[0040] S4.3. Use the calculated rate of change information and apply the GD optimization algorithm to update the execution time and resource allocation of the action. By multiplying the rate of change by the adjustment factor and subtracting this value from the current parameters, gradually optimize the action configuration to gradually reduce the value of the loss cost metric. The parameter update formula is:

[0041]

[0042]

[0043] η determines the step size of the rate of change decrease and is used to control the magnitude of each parameter update. Update the decision variables of each action according to the calculated rate of change, and the updated value moves in the direction of minimizing the loss cost metric.

[0044] S4.4. Repeat the process of calculating the rate of change and updating the action parameters until the loss cost metric reaches the set number of iterations. Obtain an optimized action execution plan to ensure the minimization of resource and time consumption, thereby improving the efficiency and effectiveness of overall action execution. Project the updated t and r back to the convex set and ensure that the action time and resource allocation values are within a reasonable range.

[0045] Further, S5 includes:

[0046] S5.1. According to the actual time and resource consumption data when the autonomous aircraft executes the action, adjust the time allocation of the action, the parameters of the loss cost metric, and the importance coefficient;

[0047] S5.2. Recalculate the loss cost metric and adjust the execution strategy of subsequent actions according to the updated loss cost metric. By optimizing the action time and resource allocation again, ensure the real-time performance and efficiency of action execution.

[0048] Further, S6 includes:

[0049] S6.1. According to the action time and resource allocation plan after multiple iterations and ensure that the plan meets all task objectives and constraints, generate the final plan. During the action execution process, continuously monitor the action execution situation. Perform closed-loop optimization according to the real-time feedback data to achieve the effect of dynamically adjusting the plan.

[0050] The present invention also provides an online convex optimization system for UAV HTN mission planning, including:

[0051] The task model construction module is used to construct the complex tasks faced by the autonomous aircraft, including task objectives, constraints, available resources, etc., and clearly define the dependencies and execution processes between tasks;

[0052] The hierarchical task decomposition module is used to decompose high-level tasks into actions, construct a hierarchical task network to handle complex tasks and generate executable plans;

[0053] The online convex optimization module is used to optimize parameters based on the plan and loss cost metrics to ensure minimizing resource and time consumption to achieve efficient action execution;

[0054] The state monitoring and feedback module is used to monitor the execution state of the autonomous aircraft actions in real time and collect feedback data to ensure that the system can obtain execution information immediately;

[0055] The dynamic adjustment and replanning module is used to dynamically adjust the action execution plan according to real-time feedback to enhance the flexibility and real-time performance of the system in a changing environment.

[0056] Advantages of the present invention:

[0057] 1). The present invention constructs an online convex optimization model for HTN task planning of autonomous aircraft, which uses online optimization technology to adjust the time and resource consumption of task execution in real time for the uncertainties and dynamic environment faced in the task planning process of autonomous aircraft, thereby improving the efficiency and real-time performance of task execution. This method establishes a task model through a planning domain language, defines the initial task decomposition structure and constraints. Using the HTN algorithm, complex tasks are decomposed into multiple levels of subtasks to obtain an executable plan. A convex set is constructed based on the time and resource consumption of tasks and the loss cost metrics are defined. Using the GD optimization algorithm, the action execution time and resource consumption are optimized online. The loss cost metrics are updated according to the actual feedback, and the execution strategy of subsequent actions is dynamically adjusted to achieve the optimal time and resources for task execution, and finally an optimal plan is obtained. Generally speaking, the present invention can effectively improve the execution efficiency of autonomous aircraft actions, reduce resource waste, and can flexibly adjust the plan according to the actual situation.

[0058] 2). The present invention is different from traditional planning methods. Traditional methods have poor dynamic performance, lack real-time evaluation of actual resources and time, and are limited in dealing with uncertain environments, making it difficult to ensure the efficiency and accuracy of action execution. In contrast, the present invention conducts dynamic intelligent planning by using an online optimization algorithm and makes real-time adjustments to the final action execution results, thereby avoiding waste of resources and time of the autonomous aircraft and improving the accuracy and reliability of action execution. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a schematic flowchart of an online convex optimization method for UAV HTN mission planning disclosed by the present invention;

[0060] Figure 2 It is a structural diagram of an online convex optimization system for UAV HTN mission planning disclosed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0061] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments in the present invention belong to the scope of protection of the present invention.

[0062] Embodiment 1

[0063] Figure 1 As shown, this embodiment provides an online convex optimization method for UAV HTN mission planning, and the specific steps are as follows:

[0064] S1. Establish a PDDL task model according to the mission objectives and constraint conditions faced by the autonomous aircraft.

[0065] S1 specifically includes the following steps:

[0066] S1.1. Model the task model based on the planning domain language, which includes all possible tasks that the autonomous aircraft can perform and the types, predicates, constraints, action constraints, and task structures required to describe them. For example, the main tasks include transport, spray pesticide, spray, patrol, and the types include uav, location, etc. The modeling content is as shown in Table 1 below. The following constructs the tasks that the autonomous aircraft needs to perform.

[0067] Table 1 Complex tasks of the autonomous aircraft and construction basis

[0068] task types predicates transport uav on_ground(uav) spraypesticide location in_air(uav) spray at_location(uav,location) patrol has_goods(uav) mission_completed(location) Watered(location) time_passed resource_used

[0069] As shown in Table 2, constraints such as time limits and resource limits are formulated according to the actions that the autonomous aircraft needs to perform. The transport task that the autonomous aircraft needs to complete needs to meet the constraints of the following four actions to complete the task. And the execution of each action should meet its relevant constraints as shown in Table 2. The condition that the action of taking off needs to meet is that it needs to land first.

[0070] Table 2 Constraints Related to the Original Tasks of the Autonomous Aircraft

[0071]

[0072] The task planning that the autonomous aircraft needs to perform needs to be carried out according to the initial state of the autonomous aircraft. Table 3 shows the current initial state and target tasks of the autonomous aircraft.

[0073] Table 3 Task Planning Objectives of the Autonomous Aircraft

[0074]

[0075] S1.2. Use the planning domain language to establish the decomposition structure of the task, define complex tasks and actions and their dependencies. Describe the resources required for each action and set its estimated execution time. As shown in Tables 1, 2, and 3, all are the contents that need to be modeled by the planning domain language. Through the traditional planning domain language, the actions and consumptions that the autonomous aircraft needs to perform are described, and finally a domain file and a problem file are generated to describe the task and formulate the goal respectively.

[0076] S2. Use the HTN algorithm to decompose the complex task to obtain the initial plan.

[0077] S2 specifically includes the following steps:

[0078] S2.1. Use the HTN algorithm to decompose the complex task into actions at multiple levels. For example, tasks such as transportation, patrol, pesticide spraying, and watering performed by the autonomous aircraft can all be decomposed into a series of actions;

[0079] S2.2. Select the decomposition path of the task according to the current complex task requirements, as shown in Table 4. Ensure that the plan meets the current task goals and constraints, as shown in Tables 2 and 3.

[0080] Generate a plan composed entirely of actions. For example, the complex task of transport can be decomposed into four actions: take off, pick up goods, transport, and land.

[0081] Table 4 Task Decomposition of Transportation Tasks

[0082]

[0083] S3. Construct a convex set to represent the resource and time constraints of the action according to the time and resources required for the action of the autonomous aircraft, and construct a loss cost index to measure the total resource consumption and time consumption of the task.

[0084] S3 specifically includes the following steps:

[0085] S3.1. Divide each action in the plan into a resource set and a time set according to its resource consumption and time consumption, construct a convex set with the data in the resource set and the time set, and define the boundary conditions of the convex set to ensure that the action is executed within the feasible region.

[0086] The convex set consists of the set of resource consumption T = {t A , t B , t C , t D} and the set of time consumption R = {r a , r b , r c , r d} of all possible actions, which is represented as a range in a multi-dimensional space, and its expression is as follows:

[0087] X = {(t A , t B , t C , t D , r a , r b , r c , r d )}

[0088] To ensure that the final optimization has a lower bound and conforms to the facts, it is necessary to set boundaries for each variable in all sets according to prior knowledge, as shown in Table 5 and Table 6.

[0089] Table 5 Task Time Limits

[0090] min time(minutes) max 5 <![CDATA[t A (take_off)]]> 12 24 <![CDATA[t B (flytopickup)]]> 38 37 <![CDATA[t C (transportGoods)]]> 52 6 <![CDATA[t D (land)]]> 13

[0091] Table 6 Task Resource Limits

[0092] min resource(units) max 3 <![CDATA[r a (take_off)]]> 8 7 <![CDATA[r b (flytopickup)]]> 13 12 <![CDATA[r c (transportGoods)]]> 20 3 <![CDATA[r d (land)]]> 6

[0093] S3.2. Use the time and resource constraints of actions in the convex set to construct a loss cost metric for actions to measure the total resource consumption and time consumption of the task. Take the loss cost metric as a quantitative goal and evaluate the execution efficiency of the autonomous aircraft under different action plans according to this loss cost metric.

[0094] The loss cost metric is a convex function that measures the cost of decision-making. The formula for the loss cost metric is as follows:

[0095] L(t, r) = ∑c i *t i + ∑a i *r i + λ∑(t j - t i ) 2

[0096] Among them, c i represents that the importance coefficient of the execution time is set to 0.65, a i represents that the importance coefficient of the resource consumption is set to 0.35, (t j - t i ) 2 represents the time dependence constraint between actions, λ is the penalty coefficient of the time difference between actions, which is set to 0.32 to control the continuity of action execution. The parameter settings are shown in Table 7. The first two weighted sums represent the resource and time consumption. The last term represents the time interval between action executions to ensure the continuity of time allocation.

[0097] Table 7 Parameter Setting Table

[0098] <![CDATA[c i > <![CDATA[a i > η λ Max_iter 0.65 0.35 0.52 0.32 100

[0099] S4. Optimize the action time and resources using the GD optimization algorithm to minimize the total time and resource consumption of action execution, thereby optimizing its loss cost metric.

[0100] S4 specifically includes the following steps:

[0101] S4.1. Set the initial parameters in the GD optimization algorithm, control the amplitude of parameter changes in each update according to the adjustment factor η, and set the condition for the algorithm to stop optimizing through the number of iterations Max_iter. Among them, η determines the step size of the change rate decrease, which is used to control the amplitude of each parameter update. By determining the appropriate η to be 0.52, the optimization path will reach the optimum quickly and stably. Max_iter limits the number of optimization iterations, which is set to 100 times.

[0102] S4.2. Calculate the partial derivatives of the loss cost metric with respect to action time and resources based on the current plan and loss cost metric to obtain the rate of change information. The rate of change reflects the sensitivity of the current action time to the support configuration. The rate of change calculation formula is as follows:

[0103]

[0104] Calculate the rate of change for each action time decision, which represents the impact of the loss cost metric of that action on the rate of change of time.

[0105]

[0106] According to the currently preset resource decision, calculate its rate of change, which describes the impact of the loss cost metric of this action on the rate of change of resources.

[0107] S4.3. Use the calculated rate of change information and apply the rate of change descent method to update the execution time and resource allocation of the action. By multiplying the rate of change by an adjustment factor and subtracting this value from the current parameters, gradually optimize the configuration of the action to gradually reduce the value of the loss cost metric. The parameter update formula is as follows:

[0108]

[0109] S4.4. Repeat the process of calculating the rate of change and updating the action parameters until the loss cost metric reaches the set number of iterations. Obtain an optimized action execution plan to ensure the minimization of resource and time consumption, thereby improving the efficiency and effectiveness of the overall action execution. Project the updated t and r back to the convex set and ensure that the action time and resource allocation values are within a reasonable range.

[0110] S5. Update the loss cost metric based on the actual feedback of the autonomous aircraft's execution and dynamically adjust the execution strategy of subsequent actions.

[0111] S5 specifically includes the following steps:

[0112] S5.1. Adjust the time allocation of the action, the parameters of the loss cost metric, and the importance coefficient according to the actual time and resource consumption data when the autonomous aircraft executes the action.

[0113] S5.2. Recalculate the loss cost metric and adjust the execution strategy of subsequent actions according to the updated loss cost metric. By optimizing the action time and resource allocation again, ensure the real-time performance and efficiency of the action execution.

[0114] S6. Generate the final plan, and perform closed-loop optimization and real-time update.

[0115] S6 specifically includes the following steps:

[0116] S6.1. Based on the action time and resource allocation plan after multiple optimizations, and ensuring that the action meets all task objectives and constraints, generate the final plan. During the execution of the action, continuously monitor the execution status of the action. Perform closed-loop optimization according to the real-time feedback data to achieve the effect of dynamically adjusting the plan.

[0117] Embodiment 2

[0118] As Figure 2 shown, this embodiment provides an online convex optimization system for UAV HTN task planning, including a task model construction module, a hierarchical task decomposition module, an online convex optimization module, a state monitoring and feedback module, and a dynamic adjustment and replanning module;

[0119] The task model construction module is used to construct the complex tasks faced by the autonomous aircraft, including task objectives, constraints, and available resources, etc., and clearly define the dependency relationships and execution processes between tasks;

[0120] The hierarchical task decomposition module is used to decompose high-level tasks into actions and construct a hierarchical task network to handle complex tasks and generate an executable plan;

[0121] The online convex optimization module is used to perform parameter optimization based on the plan and loss cost metrics to ensure minimizing resource and time consumption to achieve efficient action execution;

[0122] The state monitoring and feedback module is used to monitor the execution status of the autonomous aircraft action in real time and collect feedback data to ensure that the system can obtain execution information immediately;

[0123] The dynamic adjustment and replanning module is used to dynamically adjust the action execution plan according to the real-time feedback to enhance the flexibility and real-time performance of the system in a changing environment.

[0124] Finally, it should be noted that the above 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 with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An online convex optimization method for UAV HTN mission planning, characterized in that: The steps include: S1. Establish a PDDL task model based on the mission objectives and constraints faced by the autonomous aircraft; S2. Using the HTN planning algorithm, complex tasks are decomposed into multiple actions to generate an initial plan. S3. According to the time and resources consumed by the action of the autonomous aircraft, a convex set is constructed to represent the resource and time constraints of the action, and a loss cost indicator is constructed to measure the total resource consumption and time consumption of the action; S4. Use the GD (Gradient Descent) optimization algorithm to optimize the time and resources of the action to minimize the total time and resource consumption of the action execution, thereby optimizing its loss cost indicator; S5, update the loss cost indicator according to the actual feedback of the autonomous aircraft execution, and dynamically adjust the execution strategy of subsequent actions; S6. Generate the final plan, perform closed-loop optimization and update in real time.

2. The online convex optimization method for UAV HTN mission planning according to claim 1 is characterized in that: The S1 includes: S1.

1. Define the specific mission objectives of the autonomous aircraft according to the actions that the autonomous aircraft needs to perform, and formulate constraints such as time limits and resource limits; S1.

2. Use the planning domain language to establish a task decomposition structure, define complex tasks and primitive tasks and the dependencies between them, describe the resources required for each task and action, and set an estimated execution time for it.

3. The online convex optimization method for UAV HTN mission planning according to claim 1 is characterized in that: The S2 includes: S2.1, use HTN planning algorithm to decompose complex tasks into multiple actions; S2.

2. Select the task decomposition path according to the current complex task requirements, ensure that the plan meets the current task objectives and constraints, and generate a plan composed entirely of actions.

4. The online convex optimization method for UAV HTN mission planning according to claim 1 is characterized in that: The S3 includes: S3.

1. Divide each action in the plan into a resource set and a time set according to its resource consumption and time cost. Use the data in the resource set and the time set to construct a convex set, and define the boundary conditions of the convex set to ensure that the action is executed within the feasible domain. S3.

2. Using the time and resource constraints of the action in the convex set, we construct a loss cost index for the action to measure the total resource consumption and time consumption of the action. We use the loss cost index as a quantitative goal to evaluate the execution efficiency of the autonomous aircraft in different scenarios.

5. The online convex optimization method for UAV HTN mission planning according to claim 1 is characterized in that: The S4 includes: S4.

1. Set the initial parameters in the GD optimization algorithm, control the amplitude of parameter changes in each update according to the adjustment factor, and set the conditions for stopping the optimization of the algorithm according to the number of iterations; S4.2, based on the current plan and loss cost indicator, calculate the partial derivative of the loss cost indicator with respect to action time and resources to obtain the change rate information; S4.

3. Using the calculated change rate information, apply the GD optimization algorithm to update the execution time and resource allocation of the action. By multiplying the change rate by the adjustment factor and subtracting this value from the current parameter, the configuration of the action is gradually optimized so that the value of the loss cost indicator is gradually reduced; S4.4, repeat the process of calculating the change rate and updating the action parameters until the loss cost index reaches the set number of iterations. An optimized action execution plan is obtained to ensure the minimization of resource and time consumption, thereby improving the efficiency and effect of the overall action execution.

6. The online convex optimization method for UAV HTN mission planning according to claim 1 is characterized in that: The S5 includes: S5.

1. Adjust the time allocation of the action, the parameters of the loss cost indicator, and the importance coefficient according to the actual time and resource consumption data when the autonomous aircraft performs the action; S5.

2. Recalculate the loss cost index and adjust the execution strategy of subsequent actions according to the updated index. Ensure the real-time and efficiency of action execution by optimizing the time and resource allocation of actions again.

7. The online convex optimization method for UAV HTN mission planning according to claim 1 is characterized in that: The S6 includes: S6.

1. Generate the final execution plan based on the action time and resource allocation plan after multiple optimizations, and ensure that the plan meets all action goals and constraints. During the action execution process, continuously monitor the action execution status. Perform closed-loop optimization based on real-time feedback data to achieve the effect of dynamically adjusting the plan.

8. An online convex optimization method and system for UAV HTN mission planning according to any one of claims 1-7, characterized in that: include: Task model building module, hierarchical task decomposition module, online convex optimization module, state monitoring and feedback module, dynamic adjustment and re-planning module; The task model building module is used to build complex tasks faced by autonomous aircraft, including task objectives, constraints and available resources, and clearly define the dependencies and execution processes between tasks; The hierarchical task decomposition module is used to decompose high-level complex tasks into actions, build a hierarchical task network, process complex tasks and generate executable plans; The online convex optimization module is used to optimize parameters based on plan and loss cost indicators to ensure minimization of resource and time consumption to achieve efficient action execution; The state monitoring and feedback module is used to monitor the execution state of the autonomous aircraft action in real time and collect feedback data to ensure that the system can obtain execution information in real time; The dynamic adjustment and re-planning module is used to dynamically adjust the action execution plan according to real-time feedback to enhance the flexibility and real-time performance of the system in a changing environment.

Images