Terminal guidance resource flexible scheduling method and device

Abstract

The invention discloses a terminal guidance resource flexible scheduling method and device. The method comprises the following steps: constructing a behavior tree comprising a root node layer, a system level control node layer, a single machine level control node layer, a bottom layer condition and an action node layer from top to bottom; constructing a Bayesian network according to the situation information detected by the aircraft to be decided; the behavior tree determines an intelligent agent corresponding to the aircraft to be guided based on probability distribution of a Bayesian network aircraft implementation countering strategy; determining a guidance control constraint and a detection strategy based on the intelligent agent corresponding to the to-be-guided aircraft, and performing online trajectory planning of the to-be-guided aircraft based on the guidance control constraint; and based on the detection strategy, updating the detection action and situation information of the to-be-guided aircraft. According to the method, situation changes can be sensed more accurately, and the decision accuracy is improved.

Description

Technical Field

[0001] This invention relates to the field of detection and guidance technology, specifically to a flexible scheduling method and apparatus for terminal guidance resources. Background Technology

[0002] Due to the harsh natural environment and strong interference from other aircraft or spacecraft signals in terminal guidance scenarios, the mapping relationship from situational awareness to guidance strategy is becoming increasingly complex.

[0003] Traditional terminal guidance systems for aircraft are primarily rigidly designed, making them prone to performance degradation and inefficient resource utilization, significantly impacting the aircraft's detection and precision guidance effectiveness. While some research has addressed this issue by introducing flexible scheduling mechanisms to optimize resource allocation, most methods still rely on fixed resource allocation strategies. Existing flexible scheduling mechanisms fail to fully address resource scheduling and task allocation requirements and lack effective dynamic adjustment mechanisms. Therefore, achieving flexible resource scheduling in aircraft terminal guidance systems to improve system adaptability, detection capabilities, and guidance accuracy remains a critical issue that urgently needs to be addressed in current technological research. Summary of the Invention

[0004] In view of this, the present invention provides a flexible scheduling method and apparatus for terminal guidance resources, which can solve the technical problems of limited flexibility of aircraft during terminal guidance and the failure of detection and guidance units to work effectively together and fully utilize their respective advantages.

[0005] To solve the above-mentioned technical problems, the present invention is implemented as follows.

[0006] A flexible scheduling method for terminal guidance resources includes:

[0007] Step S1: Construct a top-down behavior tree including a root node layer, a system-level control node layer, a single-machine-level control node layer, a bottom-level condition layer, and an action node layer. Each action node at the bottom of the behavior tree corresponds to a pre-trained agent, which provides guidance and detection strategies to deal with the current situation of the aircraft. The root node of the root node layer is used to receive input containing task conditions, including system-level task conditions, single-machine-level task conditions, and bottom-level task conditions. The behavior tree matches nodes in the behavior tree based on task conditions until it matches the action node at the bottom of the behavior tree.

[0008] Step S2: Extract situation features from the situation information detected by the aircraft to be decided, determine the joint probability distribution from the situation features using the Bayesian conditional probability calculation method, and construct a Bayesian network; input the Bayesian network into the behavior tree, and the behavior tree determines the agent corresponding to the aircraft to be guided based on the probability distribution of the aircraft implementing countermeasure strategies in the Bayesian network.

[0009] Step S3: Determine guidance and control constraints and detection strategies based on the intelligent agent corresponding to the aircraft to be guided; perform online trajectory planning for the aircraft to be guided based on the guidance and control constraints; and update the detection actions and situation information of the aircraft to be guided based on the detection strategies.

[0010] Preferably, in step S1, for each agent:

[0011] The agent acquires training data, which includes historical data of the aircraft. This historical data includes the aircraft's state variables and corresponding situational information, and is related to the action nodes of the agent. The agent includes an initialization policy generation module, a policy function calculation module, a state value function calculation module, and a reward feedback module. The initialization policy generation module takes the aircraft's state variables and corresponding situational information as input, combines expert knowledge, and uses deep reinforcement learning. The function generates an initialization strategy; the strategy function calculation module is used to select an action strategy based on constraints, the state value function calculation module is used to evaluate the state, and the reward feedback module is used to provide reward information for the action strategy during the iteration process; the output of the agent is a guidance control constraint and a detection strategy, wherein the guidance control constraint is the trajectory planning constraint of the aircraft, and the detection strategy is the detection behavior of the aircraft under situational information.

[0012] Preferably, in step S1, the behavior tree includes, from top to bottom, a root node layer, a system-level control node layer, a single-machine-level control node layer, and a bottom-level condition and action node layer.

[0013] The root node layer is a layer consisting of a root node. The root node receives the Bayesian network and the relative distance between the aircraft and the interference source. The root node serves as the starting point for traversal and analyzes the probability distribution of the aircraft implementing countermeasure strategies from the Bayesian network. Based on the relative distance between the aircraft and the interference source and the probability distribution of the aircraft implementing countermeasure strategies, the system-level mission conditions of the aircraft cluster to which the aircraft belongs are determined.

[0014] The system-level control node layer includes two node layers: a first sequence selection layer and a first condition selection layer. Each node in the first sequence selection layer represents the execution order of system-level tasks, and each node in the first condition selection layer represents the execution conditions of system-level tasks. The system-level control node layer determines the nodes in the first sequence selection layer that match the system-level task conditions based on expert knowledge, and obtains each child node of that node, where the child nodes are some of the nodes in the first condition selection layer. It also determines the nodes in the first condition selection layer that match the system-level task conditions based on expert knowledge. Finally, it determines the single-aircraft-level task conditions of the aircraft based on the nodes in the first condition selection layer that match the system-level task conditions, the relative distance between the aircraft and the interference source, and the probability distribution of the aircraft implementing a countermeasure strategy.

[0015] The single-machine-level control node layer includes multiple first combination layers, each first combination layer corresponding sequentially to a stage of the single-machine-level task of the aircraft from top to bottom; each first combination layer includes two tree nodes, namely a second sequence selection layer and a second condition selection layer from top to bottom; each node of the second sequence selection layer represents the task execution order of the stage of its corresponding single-machine-level task, and each node of the second condition selection layer represents the task execution conditions of the stage of its corresponding single-machine-level task; the single-machine-level control node layer determines the nodes of the second sequence selection layer that match the stage of the single-machine-level task of the aircraft and the conditions of the single-machine-level task of the aircraft. The process involves: acquiring each child node of the node, wherein the child nodes are some nodes in the second condition selection layer; determining the nodes of the second condition selection layer that match the stage of the single-machine-level task of the aircraft and the single-machine-level task conditions of the aircraft; determining the bottom-level task conditions based on the nodes of the second condition selection layer in the bottom-most first combination layer obtained by matching, the relative distance between the aircraft and the interference source, and the probability distribution of the aircraft implementing a countermeasure strategy; wherein there are multiple bottom-level tasks and bottom-level task conditions, and the bottom-level tasks are the tasks executed at the bottom level of the aircraft, all bottom-level tasks are arranged in scheduling order, and the bottom-level task conditions are conditions that correspond one-to-one with the bottom-level tasks;

[0016] The underlying condition and action node layer comprises multiple second combination layers from top to bottom. Each second combination layer corresponds sequentially to a underlying task of the aircraft. Each second combination layer includes two tree nodes, namely a third sequence selection layer and a third condition selection layer from top to bottom. Each node in the third sequence selection layer represents the execution order of its corresponding underlying task. The third condition selection layer in the bottommost second combination layer includes multiple selection nodes and action nodes. Each action node corresponds to a selection node, and each action node corresponds to a pre-trained agent. Each selection node represents the underlying task condition of its corresponding underlying task. Each node in the third conditional selection layer of the bottom second combination layer is a selection node, and each selection node represents the bottom task condition of its corresponding bottom task; the bottom condition and action node layer determines the node of the third order selection layer that matches the execution order of the bottom task, obtains each child node of the node, and the child nodes are some nodes in the third conditional selection layer; determines the selection node of the third conditional selection layer that matches the bottom task condition corresponding to the bottom task of the aircraft; and determines the pre-trained agent corresponding to the action node determined by the action node of the selection node determined in the bottom second combination layer.

[0017] Preferably, step S2: extracting situation features from the situation information detected by the aircraft to be decided, determining the joint probability distribution from the situation features using the Bayesian conditional probability calculation method, and constructing a Bayesian network, includes:

[0018] Step S21: Obtain the current situation within the target range detected by the guided aircraft. The current situation includes multiple situation information, including interference detection perception information, space environment perception information, and sea surface environment perception information. Perform preprocessing operations on each situation information and extract situation features. Use the probability of the situation information corresponding to the situation feature as the situation feature probability.

[0019] Step S22: Using each situation feature probability as a node in the situation network, obtain the causal relationships between the nodes of the situation network; the node set of the constructed situation network is... ,in, In a situational network where causal relationships exist, directed edges are established between nodes. Nodes representing the situation network For the nodes of the situational network It has a direct impact; the joint probability distribution is determined using the Bayesian conditional probability calculation method, and the joint probability distribution is:

[0020]

[0021] in, For joint probability distribution, For nodes Conditional probability table Represents a node The set of parent nodes, where n is the number of nodes in the node set of the situational network;

[0022] Step S23: Based on the joint probability distribution, determine the conditional probability distribution of the nodes of each situation network; superimpose the conditional probability distribution of the nodes of each situation network on the situation network to form a Bayesian network.

[0023] Preferably, step S3, which involves determining guidance and control constraints and detection strategies based on the intelligent agent corresponding to the aircraft to be guided, and performing online trajectory planning for the aircraft to be guided based on the guidance and control constraints, includes:

[0024] Step S31: Determine guidance control constraints and detection strategies based on the intelligent agent corresponding to the aircraft to be guided. The guidance control constraints include the current guidance strategy and the constraint conditions of the current guidance strategy of the aircraft to be guided.

[0025] Step S32: Based on the constraints of the current guidance strategy and current control strategy of the aircraft to be guided, perform online trajectory planning for the aircraft to be guided.

[0026] A flexible scheduling device for terminal guidance resources, comprising:

[0027] Initialization module: Configured to construct a top-down behavior tree including a root node layer, a system-level control node layer, a single-machine-level control node layer, and a bottom-level condition and action node layer. Each action node at the bottom of the behavior tree corresponds to a pre-trained agent, which provides guidance and detection strategies to deal with the current situation of the aircraft. The root node of the root node layer is used to receive input containing task conditions, including system-level task conditions, single-machine-level task conditions, and bottom-level task conditions. The behavior tree matches nodes in the behavior tree based on task conditions until it matches the action node at the bottom of the behavior tree.

[0028] Matching module: Configured to extract situational features from the situational information detected by the aircraft to be decided, determine the joint probability distribution from the situational features using the Bayesian conditional probability calculation method, and construct a Bayesian network; input the Bayesian network into the behavior tree, and the behavior tree determines the agent corresponding to the aircraft to be guided based on the probability distribution of the aircraft implementing countermeasure strategies in the Bayesian network;

[0029] Planning module: Configured to determine guidance and control constraints and detection strategies based on the intelligent agent corresponding to the aircraft to be guided; perform online trajectory planning for the aircraft to be guided based on the guidance and control constraints; and update the detection actions and situation information of the aircraft to be guided based on the detection strategies.

[0030] The present invention provides a computer-readable storage medium storing a plurality of instructions; the plurality of instructions are used by a processor to load and execute the method as described above.

[0031] The present invention provides an electronic device, characterized in that the electronic device comprises:

[0032] A processor is used to execute multiple instructions;

[0033] Memory, used to store multiple instructions;

[0034] The plurality of instructions are to be stored in the memory and loaded and executed by the processor as described above.

[0035] Beneficial effects:

[0036] (1) This invention extracts and cleans environmental situation features by constructing a Bayesian network, and uses conditional probability to calculate the probability distribution of the aircraft's countermeasure strategy, which can more accurately perceive situation changes and improve decision-making accuracy.

[0037] (2) This invention addresses the problems of slow terminal guidance decision-making response and poor mission adaptability of aircraft by introducing a flexible scheduling mechanism, enabling aircraft to adapt to changing environments. Based on expert knowledge, a behavior tree structured intelligent agent decision-making model is constructed, enabling the terminal guidance system of the aircraft to implement countermeasure strategies according to the real-time situation of the aircraft, dynamically select the optimal intelligent agent, and quickly adjust the detection and guidance strategies, thereby effectively enhancing the adaptability and flexibility of the aircraft.

[0038] (3) In view of the problem that detection and guidance strategies are difficult to adaptively optimize in complex environments, the present invention adopts a data-driven detection and guidance strategy decision-making method. By sensing and analyzing the adversarial situation in real time, resources are flexibly adjusted, so as to realize the adaptive adjustment and real-time update of the strategy by the intelligent agent. This avoids the problem of resource waste or shortage under fixed allocation strategy and improves the efficiency of resource use. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the flexible scheduling method for terminal guidance resources of the present invention;

[0040] Figure 2 This is a schematic diagram of the agent decision-making model with a behavior tree structure according to the present invention;

[0041] Figure 3 This is a schematic diagram of the flexible scheduling device for terminal guidance resources of the present invention. Detailed Implementation

[0042] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0043] like Figure 1 As shown, this invention proposes a flexible scheduling method for terminal guidance resources, the method comprising:

[0044] Step S1: Construct a top-down behavior tree including a root node layer, a system-level control node layer, a single-machine-level control node layer, a bottom-level condition layer, and an action node layer. Each action node at the bottom of the behavior tree corresponds to a pre-trained agent, which provides guidance and detection strategies to deal with the current situation of the aircraft. The root node of the root node layer is used to receive input containing task conditions, including system-level task conditions, single-machine-level task conditions, and bottom-level task conditions. The behavior tree matches nodes in the behavior tree based on task conditions until it matches the action node at the bottom of the behavior tree.

[0045] Step S2: Extract situation features from the situation information detected by the aircraft to be decided, determine the joint probability distribution from the situation features using the Bayesian conditional probability calculation method, and construct a Bayesian network; input the Bayesian network into the behavior tree, and the behavior tree determines the agent corresponding to the aircraft to be guided based on the probability distribution of the aircraft implementing countermeasure strategies in the Bayesian network.

[0046] Step S3: Determine guidance and control constraints and detection strategies based on the intelligent agent corresponding to the aircraft to be guided; perform online trajectory planning for the aircraft to be guided based on the guidance and control constraints; and update the detection actions and situation information of the aircraft to be guided based on the detection strategies.

[0047] In step S1, for each agent:

[0048] The agent acquires training data, which includes historical data of the aircraft. This historical data includes the aircraft's state variables and corresponding situational information, and is related to the action nodes of the agent. The agent includes an initialization policy generation module, a policy function calculation module, a state value function calculation module, and a reward feedback module. The initialization policy generation module takes the aircraft's state variables and corresponding situational information as input, combines expert knowledge, and uses deep reinforcement learning. The function generates an initialization strategy; the strategy function calculation module is used to select an action strategy based on constraints, the state value function calculation module is used to evaluate the state, and the reward feedback module is used to provide reward information for the action strategy during the iteration process; the output of the agent is a guidance control constraint and a detection strategy, wherein the guidance control constraint is the trajectory planning constraint of the aircraft, and the detection strategy is the detection behavior of the aircraft under situational information.

[0049] During training, the agent determines the environmental state variables and corresponding aircraft actions, obtains environmental feedback, generates a reward value by the reward feedback module, updates the state value function and policy function parameters based on the reward value, and updates the action policy by the policy function calculation module. After multiple iterations, the agent is trained when the reward value converges.

[0050] In this invention, the formula for the agent generation initialization strategy is:

[0051]

[0052] in, The aircraft's state, including position, velocity, angle of attack, and roll angle. These are the actions that an intelligent agent can take, including guidance and control constraints and detection strategies; It refers to the situational environment of the aircraft. It is the action value function The parameters, It is the action value function; In a given aircraft state ,action and the aircraft situational environment The Q value.

[0053] like Figure 2 As shown, in step S1, the behavior tree includes, from top to bottom, a root node layer, a system-level control node layer, a single-machine-level control node layer, and a bottom-level condition and action node layer.

[0054] The root node layer is a layer consisting of a root node. The root node receives the Bayesian network and the relative distance between the aircraft and the interference source. The root node serves as the starting point for traversal and analyzes the probability distribution of the aircraft implementing countermeasure strategies from the Bayesian network. Based on the relative distance between the aircraft and the interference source and the probability distribution of the aircraft implementing countermeasure strategies, the system-level mission conditions of the aircraft cluster to which the aircraft belongs are determined.

[0055] The system-level control node layer includes two node layers: a first sequence selection layer and a first condition selection layer. Each node in the first sequence selection layer represents the execution order of system-level tasks, and each node in the first condition selection layer represents the execution conditions of system-level tasks. The system-level control node layer determines the nodes in the first sequence selection layer that match the system-level task conditions based on expert knowledge, and obtains each child node of that node, where the child nodes are some of the nodes in the first condition selection layer. It also determines the nodes in the first condition selection layer that match the system-level task conditions based on expert knowledge. Finally, it determines the single-aircraft-level task conditions of the aircraft based on the nodes in the first condition selection layer that match the system-level task conditions, the relative distance between the aircraft and the interference source, and the probability distribution of the aircraft implementing a countermeasure strategy.

[0056] The single-machine-level control node layer includes multiple first combination layers, each first combination layer corresponding sequentially to a stage of the single-machine-level task of the aircraft from top to bottom; each first combination layer includes two tree nodes, namely a second sequence selection layer and a second condition selection layer from top to bottom; each node of the second sequence selection layer represents the task execution order of the stage of its corresponding single-machine-level task, and each node of the second condition selection layer represents the task execution conditions of the stage of its corresponding single-machine-level task; the single-machine-level control node layer determines the nodes of the second sequence selection layer that match the stage of the single-machine-level task of the aircraft and the conditions of the single-machine-level task of the aircraft. The process involves: acquiring each child node of the node, wherein the child nodes are some nodes in the second condition selection layer; determining the nodes of the second condition selection layer that match the stage of the single-machine-level task of the aircraft and the single-machine-level task conditions of the aircraft; determining the bottom-level task conditions based on the nodes of the second condition selection layer in the bottom-most first combination layer obtained by matching, the relative distance between the aircraft and the interference source, and the probability distribution of the aircraft implementing a countermeasure strategy; wherein there are multiple bottom-level tasks and bottom-level task conditions, and the bottom-level tasks are the tasks executed at the bottom level of the aircraft, all bottom-level tasks are arranged in scheduling order, and the bottom-level task conditions are conditions that correspond one-to-one with the bottom-level tasks;

[0057] The underlying condition and action node layer comprises multiple second combination layers from top to bottom. Each second combination layer corresponds sequentially to a underlying task of the aircraft. Each second combination layer includes two tree nodes, namely a third sequence selection layer and a third condition selection layer from top to bottom. Each node in the third sequence selection layer represents the execution order of its corresponding underlying task. The third condition selection layer in the bottommost second combination layer includes multiple selection nodes and action nodes. Each action node corresponds to a selection node, and each action node corresponds to a pre-trained agent. Each selection node represents the underlying task condition of its corresponding underlying task. Each node in the third conditional selection layer of the bottom second combination layer is a selection node, and each selection node represents the bottom task condition of its corresponding bottom task; the bottom condition and action node layer determines the node of the third order selection layer that matches the execution order of the bottom task, obtains each child node of the node, and the child nodes are some nodes in the third conditional selection layer; determines the selection node of the third conditional selection layer that matches the bottom task condition corresponding to the bottom task of the aircraft; and determines the pre-trained agent corresponding to the action node determined by the action node of the selection node determined in the bottom second combination layer.

[0058] Step S2: Extracting situational features from the situational information detected by the aircraft to be decided, determining the joint probability distribution from the situational features using the Bayesian conditional probability calculation method, and constructing a Bayesian network, including:

[0059] Step S21: Obtain the current situation within the target range detected by the guided aircraft. The current situation includes multiple situation information, including interference detection perception information, space environment perception information, and sea surface environment perception information. Perform preprocessing operations on each situation information and extract situation features. Use the probability of the situation information corresponding to the situation feature as the situation feature probability.

[0060] Step S22: Using each situation feature probability as a node in the situation network, obtain the causal relationships between the nodes of the situation network; the node set of the constructed situation network is... ,in, In a situational network where causal relationships exist, directed edges are established between nodes. Nodes representing the situation network For the nodes of the situational network It has a direct impact; the joint probability distribution is determined using the Bayesian conditional probability calculation method, and the joint probability distribution is:

[0061]

[0062] in, For joint probability distribution, For nodes Conditional probability table Represents a node The set of parent nodes, where n is the number of nodes in the node set of the situational network;

[0063] Step S23: Based on the joint probability distribution, determine the conditional probability distribution of the nodes of each situation network; superimpose the conditional probability distribution of the nodes of each situation network on the situation network to form a Bayesian network.

[0064] In this invention, the preprocessing includes noise reduction, standardization, redundancy removal, and feature selection. Furthermore, the structure of the situational network can be determined using expert experience or learning data.

[0065] Step S2: Input the Bayesian network into the behavior tree. The behavior tree determines the agent corresponding to the aircraft to be guided based on the probability distribution of the aircraft implementing countermeasure strategies according to the Bayesian network, including:

[0066] Step S24: The behavior tree obtains the Bayesian network, and the weighted joint probability of the countermeasure behavior is determined based on the Bayesian network:

[0067] ;

[0068] The weighted joint probability is normalized to the interval [0,1], and the normalized weighted joint probability is... for:

[0069]

[0070] in, The weighted joint probability of the countermeasure. All of these are nodes that interfere with aircraft. Given the current situation, It is the minimum of all weighted joint probabilities. It is the maximum value among all weighted joint probabilities;

[0071] Step S25: Analyze the probability distribution of the aircraft implementing a countermeasure strategy from the normalized weighted joint probability;

[0072] Step S26: The behavior tree model determines the agent corresponding to the current situation of the aircraft to be guided based on the probability distribution of the aircraft implementing countermeasure strategies in the Bayesian network.

[0073] In this invention, the agent that determines the current state of the aircraft to be guided can be represented as:

[0074]

[0075] in, For the number of agents, Indicates the first The selection criteria for each agent.

[0076] Step S3 involves determining guidance and control constraints and detection strategies based on the intelligent agent corresponding to the aircraft to be guided, and then performing online trajectory planning for the aircraft to be guided based on the guidance and control constraints, including:

[0077] Step S31: Determine guidance control constraints and detection strategies based on the intelligent agent corresponding to the aircraft to be guided. The guidance control constraints include the current guidance strategy and the constraint conditions of the current guidance strategy of the aircraft to be guided.

[0078] Step S32: Based on the constraints of the current guidance strategy and current control strategy of the aircraft to be guided, perform online trajectory planning for the aircraft to be guided;

[0079] The mathematical model for the constraints of the current guidance strategy is as follows:

[0080]

[0081]

[0082] Assign coefficients to the objective function. For the angle of attack, The tilt angle, Let be the objective function. The longitude of the destination point for the aircraft. To determine the longitude of the desired destination, The latitude at which the aircraft reaches its destination. To determine the desired latitude at the destination, The altitude at which the aircraft reaches its destination. To the desired height at the destination, For the derivative of the state variable, The equation of motion includes state variables and unit step size. unit step length, The state matrix, The equation of motion includes the control variable and the unit step size. The equation of motion includes state variables, control variables, and unit step size. for equation, for equation, For the minimum angle of attack, For maximum angle of attack, Minimum tilt angle, The maximum yaw angle, For the rate of change of angle of attack, For the maximum angle of attack change rate, The rate of change of the tilt angle, The maximum rate of change of the tilt angle. To perform absolute value operations, This represents the heat flux value of the k-th reference trajectory after linearization. This represents the overload value of the k-th reference trajectory after linearization. This represents the dynamic pressure value of the k-th reference trajectory after linearization. To perform the partial derivative operation, This represents the heat flux density of the aircraft. For the altitude of the aircraft, This represents the aircraft's overload value. For the dynamic pressure of the aircraft, These are the discrete points after linearization. For discrete points on the k-th trajectory after linearization, For the speed of the aircraft, The altitude of the aircraft on the k-th trajectory after linearization. Let the velocity of the aircraft on the k-th trajectory be the linearized velocity. The angle of attack on the k-th trajectory after linearization. This represents the maximum value of the heat flux density. This represents the maximum value of the overload. This represents the maximum value of the dynamic pressure. and These are the longitude and latitude of the location where the aircraft to be guided is pulled off course. and These are the actual longitude and latitude of the aircraft to be guided. The longitude of the k-th trajectory after linearization. The latitude of the k-th trajectory after linearization. To determine the radius of the deflection range, For state variables, Let K be the state variable on the k-th trajectory after linearization. For trust region coefficients, These are the initial and final state vectors of the aircraft to be guided, respectively. These are the initial values ​​for the altitude, longitude, latitude, velocity, velocity tilt angle, velocity deflection angle, angle of attack, and roll angle of the aircraft to be guided. These are the terminal values ​​for the altitude, longitude, latitude, speed, speed tilt angle, speed deflection angle, angle of attack, and roll angle of the aircraft to be guided.

[0083] Online trajectory planning for the aircraft to be guided includes:

[0084] Obtain current guidance constraints:

[0085]

[0086] Using the current guidance constraints as initial conditions, the sequential solution is obtained through iterative calculation. ,Will As a result of online trajectory planning, As a sequence of control variables Indicates the number of iterations.

[0087] Step S3, based on the detection strategy, updates the detection actions and situation information of the aircraft to be guided, including:

[0088] Step S33: Obtain the detection strategy; obtain the first detection strategy corresponding to the current situation of the aircraft to be guided from the pre-established detection strategy library;

[0089] Step S34: Based on the first detection strategy and the first detection strategy, update the action and situation information of the aircraft to be guided to obtain the updated action and situation information of the aircraft.

[0090] like Figure 3 As shown, the present invention also provides a flexible scheduling device for terminal guidance resources, the device comprising:

[0091] Initialization module: Configured to construct a top-down behavior tree including a root node layer, a system-level control node layer, a single-machine-level control node layer, and a bottom-level condition and action node layer. Each action node at the bottom of the behavior tree corresponds to a pre-trained agent, which provides guidance and detection strategies to deal with the current situation of the aircraft. The root node of the root node layer is used to receive input containing task conditions, including system-level task conditions, single-machine-level task conditions, and bottom-level task conditions. The behavior tree matches nodes in the behavior tree based on task conditions until it matches the action node at the bottom of the behavior tree.

[0092] Matching module: Configured to extract situational features from the situational information detected by the aircraft to be decided, determine the joint probability distribution from the situational features using the Bayesian conditional probability calculation method, and construct a Bayesian network; input the Bayesian network into the behavior tree, and the behavior tree determines the agent corresponding to the aircraft to be guided based on the probability distribution of the aircraft implementing countermeasure strategies in the Bayesian network;

[0093] Planning module: Configured to determine guidance and control constraints and detection strategies based on the intelligent agent corresponding to the aircraft to be guided; perform online trajectory planning for the aircraft to be guided based on the guidance and control constraints; and update the detection actions and situation information of the aircraft to be guided based on the detection strategies.

[0094] The specific embodiments described above only illustrate the design principles of the present invention. The shapes and names of the components in this description may differ and are not limited. Therefore, those skilled in the art can modify or make equivalent substitutions to the technical solutions described in the foregoing embodiments; and these modifications and substitutions do not depart from the inventive spirit and technical solutions of the present invention, and should all fall within the protection scope of the present invention.

Claims

1. A flexible scheduling method for terminal guidance resources, characterized in that, The methods include: Step S1: Construct a top-down behavior tree including a root node layer, a system-level control node layer, a single-machine-level control node layer, a bottom-level condition layer, and an action node layer. Each action node at the bottom of the behavior tree corresponds to a pre-trained agent, which provides guidance and detection strategies to deal with the current situation of the aircraft. The root node of the root node layer is used to receive input containing task conditions, including system-level task conditions, single-machine-level task conditions, and bottom-level task conditions. The behavior tree matches nodes in the behavior tree based on task conditions until it matches the action node at the bottom of the behavior tree. Step S2: Extract situation features from the situation information detected by the aircraft to be decided, determine the joint probability distribution from the situation features using the Bayesian conditional probability calculation method, and construct a Bayesian network; input the Bayesian network into the behavior tree, and the behavior tree determines the agent corresponding to the aircraft to be guided based on the probability distribution of the aircraft implementing countermeasure strategies in the Bayesian network. Step S3: Determine guidance and control constraints and detection strategies based on the intelligent agent corresponding to the aircraft to be guided; perform online trajectory planning for the aircraft to be guided based on the guidance and control constraints; and update the detection actions and situation information of the aircraft to be guided based on the detection strategies.

2. The method as described in claim 1, characterized in that, In step S1, for each agent: The agent acquires training data, which includes historical data of the aircraft. This historical data includes the aircraft's state variables and corresponding situational information, and is related to the action nodes of the agent. The agent includes an initialization policy generation module, a policy function calculation module, a state value function calculation module, and a reward feedback module. The initialization policy generation module takes the aircraft's state variables and corresponding situational information as input, combines expert knowledge, and uses deep reinforcement learning. Function generation initialization strategy; The strategy function calculation module is used to select action strategies based on constraints, the state value function calculation module is used to evaluate the state, and the reward feedback module is used to provide reward information for the action strategies during the iteration process. The output of the agent is the guidance control constraints and the detection strategy. The guidance control constraints are the trajectory planning constraints of the aircraft, and the detection strategy is the detection behavior of the aircraft under situational information.

3. The method as described in claim 1, characterized in that, In step S1, the behavior tree includes, from top to bottom, a root node layer, a system-level control node layer, a single-machine-level control node layer, and a bottom-level condition and action node layer. The root node layer is a layer consisting of a root node. The root node receives the Bayesian network and the relative distance between the aircraft and the interference source. The root node serves as the starting point for traversal and analyzes the probability distribution of the aircraft implementing countermeasure strategies from the Bayesian network. Based on the relative distance between the aircraft and the interference source and the probability distribution of the aircraft implementing countermeasure strategies, the system-level mission conditions of the aircraft cluster to which the aircraft belongs are determined. The system-level control node layer includes two node layers: a first sequence selection layer and a first condition selection layer. Each node in the first sequence selection layer represents the execution order of system-level tasks, and each node in the first condition selection layer represents the execution conditions of system-level tasks. The system-level control node layer determines the nodes in the first sequence selection layer that match the system-level task conditions based on expert knowledge, and obtains each child node of that node, where the child nodes are some of the nodes in the first condition selection layer. It also determines the nodes in the first condition selection layer that match the system-level task conditions based on expert knowledge. Finally, it determines the single-aircraft-level task conditions of the aircraft based on the nodes in the first condition selection layer that match the system-level task conditions, the relative distance between the aircraft and the interference source, and the probability distribution of the aircraft implementing a countermeasure strategy. The single-machine-level control node layer includes multiple first combination layers, each first combination layer corresponding sequentially to a stage of the single-machine-level task of the aircraft from top to bottom; each first combination layer includes two tree nodes, namely a second sequence selection layer and a second condition selection layer from top to bottom; each node of the second sequence selection layer represents the task execution order of the stage of its corresponding single-machine-level task, and each node of the second condition selection layer represents the task execution conditions of the stage of its corresponding single-machine-level task; the single-machine-level control node layer determines the nodes of the second sequence selection layer that match the stage of the single-machine-level task of the aircraft and the conditions of the single-machine-level task of the aircraft. The process involves: acquiring each child node of the node, wherein the child nodes are some nodes in the second condition selection layer; determining the nodes of the second condition selection layer that match the stage of the single-machine-level task of the aircraft and the single-machine-level task conditions of the aircraft; determining the bottom-level task conditions based on the nodes of the second condition selection layer in the bottom-most first combination layer obtained by matching, the relative distance between the aircraft and the interference source, and the probability distribution of the aircraft implementing a countermeasure strategy; wherein there are multiple bottom-level tasks and bottom-level task conditions, and the bottom-level tasks are the tasks executed at the bottom level of the aircraft, all bottom-level tasks are arranged in scheduling order, and the bottom-level task conditions are conditions that correspond one-to-one with the bottom-level tasks; The underlying condition and action node layer comprises multiple second combination layers from top to bottom. Each second combination layer corresponds sequentially to a underlying task of the aircraft. Each second combination layer includes two tree nodes, namely a third sequence selection layer and a third condition selection layer from top to bottom. Each node in the third sequence selection layer represents the execution order of its corresponding underlying task. The third condition selection layer in the bottommost second combination layer includes multiple selection nodes and action nodes. Each action node corresponds to a selection node, and each action node corresponds to a pre-trained agent. Each selection node represents the underlying task condition of its corresponding underlying task. Each node in the third conditional selection layer of the bottom second combination layer is a selection node, and each selection node represents the bottom task condition of its corresponding bottom task; the bottom condition and action node layer determines the node of the third order selection layer that matches the execution order of the bottom task, obtains each child node of the node, and the child nodes are some nodes in the third conditional selection layer; determines the selection node of the third conditional selection layer that matches the bottom task condition corresponding to the bottom task of the aircraft; and determines the pre-trained agent corresponding to the action node determined by the action node of the selection node determined in the bottom second combination layer.

4. The method as described in claim 3, characterized in that, Step S2: Extracting situational features from the situational information detected by the aircraft to be decided, determining the joint probability distribution from the situational features using the Bayesian conditional probability calculation method, and constructing a Bayesian network, including: Step S21: Obtain the current situation within the target range detected by the guided aircraft. The current situation includes multiple situation information, including interference detection perception information, space environment perception information, and sea surface environment perception information. Perform preprocessing operations on each situation information and extract situation features. Use the probability of the situation information corresponding to the situation feature as the situation feature probability. Step S22: Using each situation feature probability as a node in the situation network, obtain the causal relationships between the nodes of the situation network; the node set of the constructed situation network is... ,in, In a situational network where causal relationships exist, directed edges are established between nodes. Nodes representing the situation network For the nodes of the situational network It has a direct impact; the joint probability distribution is determined using the Bayesian conditional probability calculation method, and the joint probability distribution is: in, For joint probability distribution, For nodes Conditional probability table Represents a node The set of parent nodes, where n is the number of nodes in the node set of the situational network; Step S23: Based on the joint probability distribution, determine the conditional probability distribution of the nodes of each situation network; superimpose the conditional probability distribution of the nodes of each situation network on the situation network to form a Bayesian network.

5. The method according to any one of claims 1-4, characterized in that, Step S3 involves determining guidance and control constraints and detection strategies based on the intelligent agent corresponding to the aircraft to be guided, and then performing online trajectory planning for the aircraft to be guided based on the guidance and control constraints, including: Step S31: Determine guidance control constraints and detection strategies based on the intelligent agent corresponding to the aircraft to be guided. The guidance control constraints include the current guidance strategy and the constraint conditions of the current guidance strategy of the aircraft to be guided. Step S32: Based on the constraints of the current guidance strategy and current control strategy of the aircraft to be guided, perform online trajectory planning for the aircraft to be guided.

6. A flexible scheduling device for terminal guidance resources, used to execute the method of any one of claims 1-5, characterized in that, The device includes: Initialization module: Configured to construct a top-down behavior tree including a root node layer, a system-level control node layer, a single-machine-level control node layer, and a bottom-level condition and action node layer. Each action node at the bottom of the behavior tree corresponds to a pre-trained agent, which provides guidance and detection strategies to deal with the current situation of the aircraft. The root node of the root node layer is used to receive input containing task conditions, including system-level task conditions, single-machine-level task conditions, and bottom-level task conditions. The behavior tree matches nodes in the behavior tree based on task conditions until it matches the action node at the bottom of the behavior tree. Matching module: Configured to extract situational features from the situational information detected by the aircraft to be decided, determine the joint probability distribution from the situational features using the Bayesian conditional probability calculation method, and construct a Bayesian network; input the Bayesian network into the behavior tree, and the behavior tree determines the agent corresponding to the aircraft to be guided based on the probability distribution of the aircraft implementing countermeasure strategies in the Bayesian network; Planning module: Configured to determine guidance and control constraints and detection strategies based on the intelligent agent corresponding to the aircraft to be guided; perform online trajectory planning for the aircraft to be guided based on the guidance and control constraints; and update the detection actions and situation information of the aircraft to be guided based on the detection strategies.

7. A computer-readable storage medium storing a plurality of instructions; the plurality of instructions being loaded by a processor and executing the method as claimed in any one of claims 1-5.

8. An electronic device, characterized in that, The electronic device includes: A processor is used to execute multiple instructions; Memory, used to store multiple instructions; The plurality of instructions are to be stored in the memory and loaded by the processor and executed as described in any one of claims 1-5.

Images