CN115900433A - Decision method of multi-agent unmanned countermeasure system based on SWOT analysis and behavior tree - Google Patents
Decision method of multi-agent unmanned countermeasure system based on SWOT analysis and behavior tree Download PDFInfo
- Publication number
- CN115900433A CN115900433A CN202211568902.6A CN202211568902A CN115900433A CN 115900433 A CN115900433 A CN 115900433A CN 202211568902 A CN202211568902 A CN 202211568902A CN 115900433 A CN115900433 A CN 115900433A
- Authority
- CN
- China
- Prior art keywords
- decision
- agent
- unmanned
- behavior tree
- swot
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000004458 analytical method Methods 0.000 title claims abstract description 45
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000004891 communication Methods 0.000 claims abstract description 31
- 230000006399 behavior Effects 0.000 claims description 58
- 230000009471 action Effects 0.000 claims description 31
- 238000011156 evaluation Methods 0.000 claims description 24
- 230000000875 corresponding effect Effects 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 12
- 238000010606 normalization Methods 0.000 claims description 5
- 230000003993 interaction Effects 0.000 claims description 3
- 238000004806 packaging method and process Methods 0.000 claims description 3
- 238000010845 search algorithm Methods 0.000 claims description 3
- 230000003945 visual behavior Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 238000010276 construction Methods 0.000 claims description 2
- 230000002452 interceptive effect Effects 0.000 claims description 2
- 230000003542 behavioural effect Effects 0.000 claims 6
- 238000004364 calculation method Methods 0.000 abstract description 12
- 238000013461 design Methods 0.000 abstract description 4
- 230000002159 abnormal effect Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 23
- 230000006378 damage Effects 0.000 description 6
- 230000008901 benefit Effects 0.000 description 3
- 230000007123 defense Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000004888 barrier function Effects 0.000 description 2
- 239000008280 blood Substances 0.000 description 2
- 210000004369 blood Anatomy 0.000 description 2
- 230000000295 complement effect Effects 0.000 description 2
- 230000001276 controlling effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000011273 social behavior Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a decision method of a multi-agent unmanned countermeasure system based on SWOT analysis and a behavior tree, which can be used for a real multi-agent unmanned autonomous countermeasure system, and the SWOT analysis method is used for helping the behavior tree to select a non-abnormal action type, so that the decision method is simple to realize, has good interpretability and certain self-adaptive capacity, and avoids the limitation that the traditional method needs manual design of condition nodes; a distributed communication and calculation framework is designed, decentralization is realized, calculation resources are reasonably distributed, and the application background of a real multi-agent unmanned confrontation immediate decision system in a high-dimensional state is met; a hierarchical behavior decision structure is designed, the characteristic of increasing decision precision is achieved, the autonomous behavior self-organization of the intelligent agent is facilitated, and the robustness of the real instant decision in the multi-intelligent-agent unmanned system is improved.
Description
Technical Field
The invention relates to a multi-agent instant decision-making confrontation technology, belongs to the technical field of robots and artificial intelligence, and particularly relates to a multi-agent unmanned confrontation system decision-making method based on SWOT analysis and a behavior tree.
Background
In the prior art, most decision-making technologies can only solve single agents or game environments, and cannot solve the instant decision-making task of a complex multi-agent confrontation system in the real world. Aiming at the multi-agent autonomous unmanned confrontation system, the existing decision method still has the following problems:
1. the multi-agent immediate behavior decision method is executed mostly by relying on the preset logic based on the rules, and various possible situations need to be considered carefully and a decision scheme needs to be given. For example, the history proposes a behavior control method and device for a robot based on a behavior tree in the patent "a behavior control method and device for a robot based on a behavior tree", which requires careful design of logical relationships between nodes, is time-consuming, labor-consuming, easy to consider, and prone to logic confusion. When solving the countermeasure task, in order to ensure the rigor of the logic, the structure of the behavior tree is generally very complex, and the difficulty of optimizing and testing the logic is large. The traditional method based on the behavior tree has fixed logic and poor adaptivity and flexibility.
2. The SWOT analysis method is self-management, is often applied to enterprise development strategy formulation and competitor analysis, and helps enterprises analyze own advantages and the operating environment, so that reasonable decisions are made. However, since the result obtained by the method is generally referred to by people, the decision result is not directly executable by the intelligent agent, and is difficult to be directly applied to the multi-intelligent agent unmanned confrontation system. A few methods are applied to military games, such as the research of complementary advantages of joint reconnaissance based on the SWOT analysis method proposed by Zhang Changchun in the thesis of complementary advantages of joint reconnaissance based on the SWOT analysis method, so that the joint reconnaissance can be realized. Since the method requires only two actions to be performed, the analysis result can be directly performed. However, for a complex high-dimensional multi-agent unmanned countermeasure system, two important problems are that which indexes should be selected for effective analysis and how the analysis result is interfaced with a complex action space.
3. The multi-agent autonomous countermeasure task relates to a plurality of functional modules such as target identification, autonomous positioning, attack tracking, task planning, target pose planning, path planning, attack target planning, vehicle driving posture planning, chassis control and holder control, and input state data are high-dimensional and multi-modal frequently. Most methods based on logic judgment do not comprehensively analyze various observation data information for decision making, and the optimization-based methods are difficult to model and solve in the face of complex state data structures and task requirements. The state space and the action space of the whole system are very complex, the real-time requirement of the real-time decision task on the actual robot is high, and the optimal decision scheme cannot be directly and intensively generated.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: aiming at the problems in the prior art, the decision method of the multi-agent ground unmanned countermeasure system based on SWOT analysis and behavior trees is provided, and the method is a distributed autonomous decision method of the multi-agent ground unmanned countermeasure system, which is easy to realize, debug, expand and have strong flexibility.
In order to solve the technical problems, the technical scheme provided by the invention is as follows:
a multi-agent unmanned countermeasure system decision method based on SWOT analysis and behavior tree comprises the following steps:
step 1: building a distributed communication network, and acquiring communication data in real time according to the built distributed communication network, wherein the communication data comprises interaction state data among multiple agents in the same team, countermeasure system state information and enemy state data obtained by observation; the interactive state data among the multiple agents in the same team comprises planning state result data;
step 2: building a distributed state blackboard, and inputting the communication data obtained in real time in the step 1 into the built distributed state blackboard;
and step 3: the method for constructing the behavior tree comprises the following specific steps: automatically generating a json file through an xml-based visual behavior tree editor, and analyzing the generated json file to complete the construction of a behavior tree;
and 4, step 4: constructing a SWOT situation analysis module, wherein the constructed SWOT situation analysis module comprises a situation evaluation function and a combat capability evaluation function, the situation evaluation function is used for measuring the win-or-lose trend, and the combat capability evaluation function is used for measuring the combat capability;
and 5: processing the communication data acquired in real time in the step 1 to obtain a communication data processing result, and inputting the obtained communication data processing result into the distributed state blackboard written with the communication data in the step 2, wherein the communication data processing result comprises a universal cost map set and a fighting force value of each intelligent agent;
and 6: executing the behavior tree constructed in the step 3 by using the confrontation system state information in the distributed state blackboard in the step 5;
and 7: classifying the current confrontation situation into an ST state, an SO state, a WT state and a WO state by using a situation evaluation function and a combat power evaluation function in the SWOT situation analysis module constructed in the step 4, inputting the current confrontation situation into a distributed state blackboard in a classified mode to serve as a situation state data value, and selecting action nodes in a behavior tree according to the confrontation situation in a classified mode;
and 8: and (4) calculating a decision scheme of the position of the moving target of the intelligent agent according to the action node in the behavior tree selected in the step (7), wherein the calculation method specifically comprises the following steps: distributing weight coefficients corresponding to the actions to each layer of cost map according to the general cost map set in the step 5 and the action node types selected in the step 7, carrying out weighted summation on each layer of cost map to obtain a target position cost map of the corresponding action, searching a grid point with the minimum cost through a heuristic search algorithm, carrying out coordinate system conversion, and obtaining a decision scheme of the moving target position of the intelligent body;
and step 9: calculating a decision scheme of a hitting target, a running attitude and a target attitude angle in the action nodes according to the action nodes in the behavior tree selected in the step 7, wherein the specific calculation method comprises the following steps:
(1) Constructing an attack target cost function according to the distance index and the fighting force value of each intelligent agent in the step 5, and calculating the id of a target attack object meeting the consistency condition according to the constructed attack target cost function;
(2) Constructing a vehicle attitude decision behavior tree, and calculating running attitude decision schemes of vehicles in different action types according to the constructed vehicle attitude decision behavior tree;
(3) Calculating a vehicle holder attitude angle decision scheme according to the enemy state data in the step 5;
step 10: sending the decision-making scheme in the step 8 to a planner and sending the decision-making scheme in the step 9 to the planner or a controller; the planner and the controller ensure that the bottom actuator can execute decision instructions, and after the planner tries to complete a planning task, the planner transmits planning state result data through the communication network in the step 1 and updates the planning state result data in the distributed blackboard in the step 5;
the decision results in the step 8 and the step 9 refer to a target motion position decision scheme of the intelligent agent in the step 8 and a target hitting, running attitude and target attitude angle decision scheme in the step 9.
In step 5, the method for processing the communication data acquired in step 1 in real time to acquire the universal cost map set includes:
(1) The grid map in the intelligent visual field range is traversed at one time through a breadth first algorithm, and a plurality of required index cost maps are generated at the same time, wherein the required index cost maps specifically comprise a distance cost map containing barrier information, a distance cost map between teammates and friends, an enemy distance cost map, a shooting cost map, the in-team concentration degree, the enemy concentration degree, the relative enemy orientation cost, an included angle surrounding the enemy, occupation cost, accessibility and other index costs;
(2) Carrying out normalization processing on each map;
(3) And (3) packaging all the cost maps, updating the universal cost map set of the distributed state blackboard, and updating the corresponding universal cost map moment set in the distributed state blackboard set in the step (2).
The invention has the following beneficial effects:
(1) The invention uses the SWOT situation analysis method to help the behavior tree to select the behavior type, visually displays the current confrontation situation, has strong interpretability, generates a decision scheme with certain self-adaptability and flexibility, and solves the problems that the traditional behavior tree completely depends on artificial logic analysis, and is easy to cause unreasonable selection of action conditions, even condition omission and the like. The problem that the SWOT analysis result is too rough and difficult to operate is solved by skillfully designing and utilizing the performability of the behavior tree.
(2) The distributed computing and communication structure is used, centralized computing is avoided, the problem of large state space dimension of the multi-agent unmanned countermeasure system can be effectively solved, computing resources of all agent computing platforms are reasonably utilized, and the requirements of real application scenes are met.
(3) The hierarchical behavior decision structure is used, the behavior decision structure is ordered, the decision precision is from coarse to fine, the intelligence degree is gradually reduced, the multi-agent autonomously realizes decision behavior self-organization, autonomous confrontation decision tasks can be realized on a real multi-agent unmanned confrontation system, and the realization method is simple and strong in robustness.
(4) A multi-agent unmanned countermeasure immediate decision-making system and method based on SWOT analysis and behavior trees are characterized in that a hierarchical immediate decision-making system is built, each agent builds a blackboard from input observation state data obtained by a distributed communication network in real time, behavior trees are built according to behavior tree logic files preset by a user, a situation analysis module based on the SWOT analysis method is built, non-fault action node types are selected, and then action nodes are executed through a heuristic optimization algorithm to obtain an executable decision-making scheme.
(5) The invention discloses a decision method of a multi-agent unmanned confrontation system based on SWOT analysis and a behavior tree, which can be used for a real multi-agent unmanned autonomous confrontation system, and the SWOT analysis method is used for helping the behavior tree to select a non-abnormal action type, so that the method is simple to realize, has good interpretability and certain self-adaptive capacity, and avoids the limitation that the traditional method needs manual design of condition nodes; a distributed communication and calculation framework is designed, decentralization is realized, calculation resources are reasonably distributed, and the application background of a real multi-agent unmanned confrontation immediate decision system in a high-dimensional state is met; a hierarchical behavior decision structure is designed, the characteristic of increasing decision precision is achieved, the autonomous behavior self-organization of the intelligent agent is facilitated, and the robustness of the real instant decision in the multi-intelligent-agent unmanned system is improved.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a block diagram of a SWOT based behavior tree of the present invention;
FIG. 3 is a schematic SWOT analysis coordinate diagram of the present invention;
FIG. 4 is a diagram of a vehicle attitude decision behavior tree according to the present invention.
Detailed Description
The invention is described in detail below with reference to the accompanying drawings and by way of example.
As shown in FIG. 1, the invention provides a multi-agent unmanned confrontation immediate decision system and method based on SWOT analysis and behavior tree, which specifically comprises the following steps:
step 1: building a distributed communication network, and acquiring interaction state data among multiple intelligent agents in the same team, confrontation system state information and observation information of the opponent and the enemy of the team according to the built distributed communication network;
step 2: building a distributed state blackboard, wherein the built distributed state blackboard is used for integrating state data of different sources in the step one and processing results of the data; the blackboard integrates communication data of all parties and state data of the communication data of all parties after processing, wherein the communication data of all parties specifically comprises: sensors such as laser radars, cameras and mileometers detect data, confrontation system data and other teammates communicate and transmit data. The processed state data includes: a general cost map set, a planner state, a blood dropping speed, a bullet consumption speed, an action selection type, a decision result, a situation state and the like. Each agent has 3 blackboards: a basic blackboard for recording the confrontation state information; the system comprises an expansion blackboard for recording the state information of the intelligent agent and observed enemy information, and an array blackboard for recording the state information and decision results of other intelligent agents in the same team;
and step 3: and constructing a behavior tree, and automatically generating a json file through an xml-based visual behavior tree editor, wherein the data structure comprises a behavior tree node type, a node name, a node type, a node id, a father node id of a node, a child node id of the node, a node characteristic variable and related description information. And loading and analyzing the json file by a program, automatically calling a library function according to the analyzed logic relation, and constructing a behavior tree corresponding to the logic. The constructed behavior tree structure is shown in FIG. 2;
and 4, step 4: constructing an SWOT situation analysis module, and respectively constructing a situation evaluation function and a fighting capacity evaluation function which are used for measuring the win-win or loss trend and the fighting capacity in the SWOT situation analysis module; in the SWOT analysis method, S (strenggths) is dominant, W (weaknesses) is disadvantageous, O (opportunities), and T (threats) are threats; as shown in fig. 3, index functions capable of judging win-or-loss and measuring the battle force are respectively designed according to the confrontation rules, the abscissa represents the relative good and bad situation (WO), and the relative good and bad situation is measured by the total damage amount scores of each party of the confrontation system; the ordinate represents the respective party competencies (ST), measured by the team's multidimensional combat assessment; the state of the abscissa (WO) is evaluated by constructing a situational evaluation function, the situational evaluation function value is equal to the sum of the own-party damage amount and the enemy damage amount, and the data is derived from the state quantity data of the countermeasure system in the distributed blackboard constructed in the step 2; the state of the ordinate (WO) is evaluated through a combat power evaluation function, wherein combat power evaluation indexes are measured by indexes such as attack accuracy, defense capability, movement capability, attack capability and stability of each intelligent agent. Wherein the hitting accuracy is measured by dividing the total damage amount of the enemy caused by the own intelligent agent by the consumed bullet amount of the own intelligent agent; the defense capacity is measured by dividing the total damage amount of the enemy agent to the enemy agent by the consumed bullet amount of the enemy; the moving capability is measured by a nonlinear transformation function of a moving speed sequence; the attack ability is measured by the number of live agents, the total blood volume and the total bullet volume of each party; stability is measured by the total injury to the team due to non-bullet shooting. After normalization, weighting and summing each evaluation index to obtain a team fighting capacity evaluation function;
and 5: and (3) updating various state data in the distributed blackboard in the step (2), acquiring real-time state data by using the distributed communication network in the step (1), and updating corresponding state data in the distributed blackboard set up in the step (2). Calculating a general cost map set according to state data in the existing blackboard, wherein the specific calculation steps are as follows:
(1) Traversing a grid map in an intelligent visual field range through a breadth first algorithm, simultaneously generating a plurality of general cost maps at one time, wherein the cost maps specifically comprise distance cost maps containing barrier information, distance cost maps between teammates and friends, enemy distance cost maps, shooting cost maps, in-team concentration degree, enemy concentration degree, relative enemy orientation cost, included angle enclosing enemy, occupation cost, accessibility and other cost maps required by different action nodes;
(2) Carrying out normalization processing on each map;
(3) Packaging all the cost maps, updating the universal cost map matrix variables of the blackboard, and updating the corresponding universal cost map matrix data in the blackboard constructed in the step 2;
step 6: operating a SWOT situation analysis module, respectively forming a situation evaluation function value and a fighting capacity evaluation function value in the step 4 according to the latest state data in the step 5, wherein the two function values form a two-dimensional coordinate, and classifying the confrontation situation state into four states of ST, SO, WT and WO according to the condition that the coordinate position falls in the interval of a two-dimensional Cartesian coordinate system; if the state is in the SO quadrant, the situation tends to win, and the attack ability of the party is stronger, the battle should be continued; if the state is in the ST quadrant, which indicates that the situation is behind, but the attack ability of the party is strong, the attack output should be strengthened as soon as possible; if the state is in the WT quadrant indicating that the situation is behind and the state is poor, the state should be replenished and attacked as soon as possible; if the state is in the WO quadrant, the situation is in windward state, but the state capability is poor, so the situation should be kept, unnecessary combat loss is avoided as much as possible, and the state is supplemented; changing the state value of an action node according to the preliminary decision task type of the situation position of the team; classifying the current confrontation situation into four states of ST, SO, WT and WO, and updating the situation state data value on the distributed blackboard in the step 2;
and 7: running the behavior tree constructed in the step 3, executing corresponding condition nodes according to the running state values of the countermeasure system in the distributed blackboard in the step 5, and calculating action type decision results according to the situation state results calculated in the step 6 under the condition of starting countermeasure;
and 8: calculating a target motion position decision scheme of an intelligent body in an action node, distributing weight coefficients corresponding to actions for each layer of cost map according to the cost map set constructed in the step 5 and the action node selected in the step 7, obtaining a target position cost map of the corresponding action after weighting and summing the cost maps of the layers according to the formula (1), searching a grid point with the minimum cost in a observable range through a heuristic search algorithm, converting a coordinate system, and obtaining a decision result of a motion target position of the intelligent body, wherein v m Representing cost maps of layers, w m For each layer of cost map weight coefficient, f (-) is a map normalization calculation operator, p u As coordinates of our robot, p e As coordinates of the enemy robot,
for the distance containing the Obstacle information, SW (-) is an intra-class divergence calculation operator, SB (-) is an inter-class divergence calculation operator, is _ shot (-) is a shot-to-shot judging function, theta (-) is an attitude angle difference calculation operator, and Obstacle (-) is an Obstacle occupancy calculation operator;
v 6 ＝SW(p i,j ,p u )
v 7 ＝SB(p i,j ,p u ,p e )
v 8 ＝is_shoot(p i,j ,p e )
v 10 ＝θ(p i,j ,p g ,p u )
v 11 ＝Obstacle(p i,j ,p g )
and step 9: calculating a striking target, an operation attitude and a target attitude angle decision scheme in the action node; in order to calculate the attack target decision result, according to the fighting force value of each enemy intelligent body in the fighting force evaluation obtained in the step 5 and the comprehensive distance index, an attack target cost function is constructed, a task commander mechanism is designed, and the id of a target attack object meeting the consistency condition is calculated; the task commander mechanism is used for ensuring that multiple agents with the same cooperative behavior attack the same enemy target together, and the specific design method comprises the following steps: and selecting a command of the collaboration task, wherein the command takes the place of the intelligent agent with the highest tactical assessment of the party when the collaboration task is executed from the first time. According to the fighting force value of each enemy intelligent body in the fighting capacity evaluation obtained in the step 5 and the comprehensive distance index, constructing an attack target cost function according to a formula (2), selecting a target enemy which is relatively close and easy to attack as an attack object, and setting the attack targets of other intelligent bodies which have the same action with the commander as the attack target id selected by the commander;
in order to calculate a vehicle attitude decision result, various vehicle attitudes such as mobile attack, swing defense, reconnaissance and full-speed movement are designed, and the vehicle attitudes are switched in real time through a behavior tree; in order to calculate a vehicle attitude decision result, a vehicle attitude decision behavior tree is constructed, and as shown in fig. 4, a vehicle attitude decision behavior tree structure is provided, so that the real-time switching of vehicle attitudes under different action types is realized; according to the enemy state data in the step 5, the vehicle holder target attitude angle always points to the position of the enemy target, and a target holder attitude angle decision result is obtained by calculating the included angle between the vehicle holder target attitude angle and the target enemy agent connection line;
step 10: and sending decision instructions to the planner and the controller. According to the step 9 and the step 10, a decision result is obtained: target pose, cradle head pose angle, vehicle pose, hitting enemy target id; sending target pose and vehicle attitude instructions to a planner, calculating an occupancy map according to the acquired data of the laser radar by the planner, performing global planning and local planning, generating a speed position state sequence, sending the speed position state sequence to a bottom controller, and controlling the vehicle to run by the controller by tracking an expected track; sending the command of hitting the enemy and the view field orientation to the cradle head and the shooting controller, and controlling the rotation angle and the muzzle of the cradle head to hit; and after the planner tries to complete the planning task, transmitting planning state result data through the communication network in the step 1, and updating the planning state result data in the distributed blackboard in the step 5.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the scope of the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A multi-agent unmanned countermeasure system decision method based on SWOT analysis and behavior tree is characterized by comprising the following steps:
step 1: building a distributed communication network, and acquiring communication data in real time according to the built distributed communication network;
step 2: building a distributed state blackboard, and inputting the communication data obtained in real time in the step 1 into the built distributed state blackboard;
and step 3: constructing a behavior tree;
and 4, step 4: constructing an SWOT situation analysis module;
and 5: updating the processing result of the communication data acquired in real time into the distributed state blackboard in the step 2;
step 6: executing the behavior tree constructed in the step 3;
and 7: classifying the current confrontation situation by using the SWOT situation analysis module constructed in the step 4, updating the classification result of the current confrontation situation to a distributed state blackboard, and selecting action nodes in a behavior tree according to the classification result of the confrontation situation;
and step 8: calculating a decision scheme of the moving target position of the intelligent agent according to the action nodes in the behavior tree selected in the step 7;
and step 9: calculating a striking target, a running attitude and a target attitude angle decision scheme in the action nodes according to the action nodes in the behavior tree selected in the step 7;
step 10: and (4) sending the decision scheme in the step (8) and the decision scheme in the step (9) to a bottom controller and a planner and updating a distributed state blackboard to finish the decision of the multi-agent unmanned countermeasure system based on the SWOT analysis and the behavior tree.
2. The decision-making method of a multi-agent unmanned countermeasure system based on SWOT analysis and behavioral trees as claimed in claim 1, wherein:
in the step 1, the communication data comprises interaction state data among multiple agents in the same team, countermeasure system state information and enemy state data obtained by observation; the interactive state data among the multiple agents in the same team comprises planning state result data.
3. The decision-making method of the multi-agent unmanned countermeasure system based on the SWOT analysis and the behavior tree as claimed in claim 1 or 2, wherein:
in the step 3, the specific method for constructing the behavior tree is as follows: and automatically generating a json file through an xml-based visual behavior tree editor, and analyzing the generated json file to complete the construction of the behavior tree.
4. The decision-making method of a multi-agent unmanned countermeasure system based on SWOT analysis and behavioral trees as claimed in claim 3, wherein:
in the step 4, the constructed SWOT situation analysis module comprises a situation evaluation function and a combat capability evaluation function, wherein the situation evaluation function is used for measuring the win-or-lose trend, and the combat capability evaluation function is used for measuring the combat capability.
5. The decision-making method of the multi-agent unmanned countermeasure system based on SWOT analysis and behavior trees as claimed in claim 4, wherein:
in the step 5, the communication data processing result comprises a general cost map set and a fighting force value of each intelligent agent.
6. The decision-making method of a multi-agent unmanned countermeasure system based on SWOT analysis and behavioral trees as claimed in claim 5, wherein:
in the step 7, the situation evaluation function and the fighting evaluation function are used to classify the current confrontation situation into four states of ST, SO, WT and WO.
7. The decision-making method of a multi-agent unmanned confrontation system based on SWOT analysis and behavior tree as claimed in claim 6, wherein:
in step 8, the specific method for calculating the decision scheme of the moving target position of the intelligent agent comprises the following steps: according to the general cost map set and the selected action node type, distributing weight coefficients corresponding to actions for each layer of cost map, carrying out weighted summation on each layer of cost map to obtain a target position cost map of the corresponding action, searching a grid point with the minimum cost through a heuristic search algorithm, and carrying out coordinate system conversion to obtain a decision scheme of the moving target position of the intelligent body.
8. The decision-making method of a multi-agent unmanned countermeasure system based on SWOT analysis and behavioral trees as claimed in claim 7, wherein:
in step 9, the decision scheme for calculating the attack target, the operation attitude and the target attitude angle in the action node is specifically as follows:
according to the distance indexes and the fighting force values of all the intelligent bodies, an attack target cost function is built, and the id of a target attack object meeting the consistency condition is calculated according to the built attack target cost function;
constructing a vehicle attitude decision behavior tree, and calculating running attitude decision schemes of vehicles in different action types according to the constructed vehicle attitude decision behavior tree;
and calculating a vehicle holder attitude angle decision scheme according to the enemy state data.
9. The decision-making method of a multi-agent unmanned countermeasure system based on SWOT analysis and behavioral trees as claimed in claim 1, wherein:
in the step 10, the decision-making scheme in the step 8 is sent to a planner, and the decision-making scheme in the step 9 is sent to the planner or a controller; the planner and the controller ensure that the bottom executor can execute decision instructions, and after the planner tries to complete a planning task, the planner transmits planning state result data through a communication network and updates the distributed state blackboard.
10. The decision-making method of a multi-agent unmanned countermeasure system based on SWOT analysis and behavioral trees as claimed in claim 1, wherein:
in step 5, the method for obtaining the universal cost map set includes:
(1) The grid map in the intelligent stereoscopic field range is traversed at one time through a breadth-first algorithm, and a plurality of required index cost maps are generated at the same time;
(2) Carrying out normalization processing on each map;
(3) And packaging all the cost maps into a map set, updating the universal cost map set of the distributed state blackboard, and updating the corresponding universal cost map set in the built distributed state blackboard.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211568902.6A CN115900433B (en) | 2022-12-08 | 2022-12-08 | Decision method of multi-agent unmanned countermeasure system based on SWOT analysis and behavior tree |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211568902.6A CN115900433B (en) | 2022-12-08 | 2022-12-08 | Decision method of multi-agent unmanned countermeasure system based on SWOT analysis and behavior tree |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115900433A true CN115900433A (en) | 2023-04-04 |
| CN115900433B CN115900433B (en) | 2024-08-13 |
Family
ID=86476413
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211568902.6A Active CN115900433B (en) | 2022-12-08 | 2022-12-08 | Decision method of multi-agent unmanned countermeasure system based on SWOT analysis and behavior tree |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115900433B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117369521A (en) * | 2023-12-04 | 2024-01-09 | 中国科学院自动化研究所 | Method, device and equipment for generating behavior tree model path for unmanned aerial vehicle decision |
| CN117669710A (en) * | 2024-02-01 | 2024-03-08 | 中国人民解放军军事科学院国防科技创新研究院 | Multi-behavior tree decision scheme aggregation method and device for game countermeasure task |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20160080351A (en) * | 2014-12-29 | 2016-07-08 | 동서대학교산학협력단 | NPC decision making modelling method in dynamic terrain analysis by prospect theory |
| CN106021816A (en) * | 2016-06-20 | 2016-10-12 | 华北电力大学（保定） | Method for achieving distributed system behavior simulated analysis tool based on behavior tree |
| CN112307622A (en) * | 2020-10-30 | 2021-02-02 | 中国兵器科学研究院 | Autonomous planning system and planning method for generating military forces by computer |
| CN113420462A (en) * | 2021-07-20 | 2021-09-21 | 西安九思智汇信息科技有限公司 | Intelligent target simulation system based on behavior tree |
| US20220244708A1 (en) * | 2021-02-04 | 2022-08-04 | Magazino Gmbh | Controlling an apparatus with a behavior tree |
-
2022
- 2022-12-08 CN CN202211568902.6A patent/CN115900433B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20160080351A (en) * | 2014-12-29 | 2016-07-08 | 동서대학교산학협력단 | NPC decision making modelling method in dynamic terrain analysis by prospect theory |
| CN106021816A (en) * | 2016-06-20 | 2016-10-12 | 华北电力大学（保定） | Method for achieving distributed system behavior simulated analysis tool based on behavior tree |
| CN112307622A (en) * | 2020-10-30 | 2021-02-02 | 中国兵器科学研究院 | Autonomous planning system and planning method for generating military forces by computer |
| US20220244708A1 (en) * | 2021-02-04 | 2022-08-04 | Magazino Gmbh | Controlling an apparatus with a behavior tree |
| CN113420462A (en) * | 2021-07-20 | 2021-09-21 | 西安九思智汇信息科技有限公司 | Intelligent target simulation system based on behavior tree |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117369521A (en) * | 2023-12-04 | 2024-01-09 | 中国科学院自动化研究所 | Method, device and equipment for generating behavior tree model path for unmanned aerial vehicle decision |
| CN117369521B (en) * | 2023-12-04 | 2024-02-13 | 中国科学院自动化研究所 | Method, device and equipment for generating behavior tree model path for unmanned aerial vehicle decision |
| CN117669710A (en) * | 2024-02-01 | 2024-03-08 | 中国人民解放军军事科学院国防科技创新研究院 | Multi-behavior tree decision scheme aggregation method and device for game countermeasure task |
| CN117669710B (en) * | 2024-02-01 | 2024-05-10 | 中国人民解放军军事科学院国防科技创新研究院 | Multi-behavior tree decision scheme aggregation method and device for game countermeasure task |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115900433B (en) | 2024-08-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Lin et al. | A Novel Improved Bat Algorithm in UAV Path Planning. | |
| CN115900433B (en) | Decision method of multi-agent unmanned countermeasure system based on SWOT analysis and behavior tree | |
| Lamont et al. | UAV swarm mission planning and routing using multi-objective evolutionary algorithms | |
| Vidal et al. | Probabilistic pursuit-evasion games: theory, implementation, and experimental evaluation | |
| CN101964019A (en) | Against behavior modeling simulation platform and method based on Agent technology | |
| CN113791634A (en) | Multi-aircraft air combat decision method based on multi-agent reinforcement learning | |
| CN116661503B (en) | Cluster track automatic planning method based on multi-agent safety reinforcement learning | |
| Khalil et al. | FED-UP: Federated deep reinforcement learning-based UAV path planning against hostile defense system | |
| Blais et al. | Reinforcement learning for swarm robotics: An overview of applications, algorithms and simulators | |
| CN114489144A (en) | Unmanned aerial vehicle autonomous maneuver decision method and device and unmanned aerial vehicle | |
| CN112651486A (en) | Method for improving convergence rate of MADDPG algorithm and application thereof | |
| CN117313561B (en) | Unmanned aerial vehicle intelligent decision model training method and unmanned aerial vehicle intelligent decision method | |
| Tolt et al. | Multi-aspect path planning for enhanced ground combat simulation | |
| Vicerra et al. | A multiple level MIMO fuzzy logic based intelligence for multiple agent cooperative robot system | |
| CN116360483A (en) | Multi-unmanned aerial vehicle cooperative attack and defense countermeasure method and system based on local observation information | |
| Zhang et al. | Multi-robot cooperation strategy in game environment using deep reinforcement learning | |
| Boudjellel et al. | Two-level Approach for Heterogeneous Multi-Robot Optimal Navigation | |
| Li et al. | Fuzzy Self-Adaptive Soccer Robot Behavior Decision System Design through ROS. | |
| Wang et al. | Learning to navigate for mobile robot with continual reinforcement learning | |
| Yang | Cooperative search by uninhabited air vehicles in dynamic environment | |
| Dawn et al. | Agent-based learning for auto-navigation within the virtual city | |
| Yepez et al. | Efficient Swarm Neutralization in Complex Environments using Multi-Agent Deep Reinforcement Learning and Adaptive Navigation Strategies | |
| Wu et al. | The collaborative strategy of multiple USVs with deep reinforcement learning method | |
| Cui et al. | Mobile robot sequential decision making using a deep reinforcement learning hyper-heuristic approach | |
| Hynes | Multi-agent simulations (mas) for assessing massive sensor coverage and deployment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |