Emergency decision-making simulation system based on behavior tree and multi-resolution disaggregation

Abstract

The present application relates to the field of simulated emergency drills, and specifically discloses an emergency decision simulation system based on behavior trees and multi-resolution deaggregation. The emergency decision simulation system based on behavior trees and multi-resolution deaggregation includes a decision simulation system and a simulation deduction system; the decision simulation system includes a decision rule editor and a decision scheduling engine; the decision rule editor includes a mission layer, a task layer, and an action layer; the mission layer behavior tree, the task layer behavior tree, and the action layer behavior tree are linked through reference nodes; the mission layer behavior tree uses a prevention and control reference node, a search and rescue reference node, an emergency reference node, and a transfer reference node as the lowest level node of the mission layer behavior tree and the root node of the task layer behavior tree, and is provided with a situation processing node and a blackboard decoration node, so that the resolution gap and its impact on simulated emergency drills can be reduced, and the flexibility of simulated emergency drills and their coordination with real-time environments can be improved.

Description

Technical Field

[0001] The present application relates to the field of simulated emergency drills, and in particular to an emergency decision-making simulation system based on behavior trees and multi-resolution deaggregation. Background Art

[0002] Emergency drill modeling and simulation include physical behavior modeling and emergency response decision modeling. Physical behavior modeling is based on physical mechanisms to model the emergency behavior and process of emergency drill entities; emergency response decision modeling is based on the objects of emergency events, emergency response resources and environmental information, to evaluate the emergency response situation, determine emergency response goals, formulate drill methods and styles, and reasonably allocate emergency resources, monitor and adjust the response action plan, and realize all decision-making issues related to emergency response. Emergency response decision modeling is mainly used to describe the emergency decision-making process of the emergency response unit. It is the core of emergency drill simulation. Its correctness and rationality have a decisive influence on the results of emergency drill simulation.

[0003] Existing simulation decision modeling mainly includes autonomous decision-making within the model and autonomous decision-making outside the model. Autonomous decision-making within the model relies on the decision logic inside the physical model, and adopts a "hard-coded" method to solidify the autonomous behavior rules of the entity inside the physical model, directly realizing the understanding of the emergency situation and forming corresponding decision results, such as decision modeling based on multiple roles (Agent). Autonomous decision-making outside the model is to manually input command decision-making knowledge with the support of expert knowledge outside the model system, give "human knowledge" to the model, and support the decision-making process of the model in simulation operation, such as finite state machine, rule-based decision modeling. In emergency simulation, the emergency decision-making process is not only restricted by deduction rules and disposal principles, but also by factors such as uncertainty in the environment, dynamic changes in the emergency situation, and limited resources. Therefore, how to make the emergency disposal decision model better adapt to complex environments and improve its flexibility and adaptability of decision-making has become the core issue of emergency decision deduction modeling.

[0004] In emergency drill simulation, how the simulation entity makes decisions based on the current situation to better cope with various challenges and improve its comprehensive effectiveness in complex emergency environments is an important issue. Current simulation decision modeling technologies mainly include finite state machine method, data-driven, agent modeling and other methods. However, the complex and diverse environment makes the simulation emergency decision solution face huge challenges. Specifically, at present, the research process of simulation emergency decision deduction modeling faces the following difficulties:

[0005] (1) Traditional simulation decision-making research is mainly conducted from two aspects: autonomous decision-making inside and outside the model. Internal decision-making relies on the decision-making logic solidified in the model to achieve an understanding of the emergency situation, and external decision-making realizes the simulation decision-making process by entering expert knowledge and experience. However, internal decision-making is difficult to develop, and it is difficult to combine expert knowledge and experience with the model. Although external decision-making uses expert experience, it is often based on finite state machines to implement rule-based decision-making. When the scale of the state space increases, it is difficult for finite state machines to maintain complex rules and condition description spaces. Behavior trees are an improvement on finite state machines, which support the addition of single behavior nodes or behavior subtrees. When faced with complex emergency response scenarios, the constructed behavior trees have problems such as poor readability, strong coupling between different modules, and fixed structures that are difficult to adapt to situation changes.

[0006] (2) The resolution of the models in the current mainstream simulation system is generally at the platform level. However, in the emergency response tasks issued by the emergency commander, the execution unit is large in scale, and the resolution of the corresponding simulation system entity model is low, resulting in a resolution gap between the emergency response plan issued by the high-level emergency commander and the simulation system model instructions. In addition, the emergency response plan has multiple levels of mission, task, and action, which makes the plan have multiple resolutions, increasing the resolution level of this difference.

[0007] (3) The variability of the emergency response environment, the diversity of emergencies, the uncertainty of the strategies for the objects of emergency response, and the existence of multi-level decision-making processes require the emergency response decision-making model to make highly complex decisions in real-time situations. Therefore, it is very cumbersome and difficult to construct a mapping from the current environment to goal-oriented actions. Summary of the invention

[0008] One advantage of the present application is that it provides an emergency decision-making simulation system based on behavior trees and multi-resolution deaggregation, wherein the emergency decision-making simulation system based on behavior trees and multi-resolution deaggregation can deaggregate low-resolution emergency tasks into high-resolution model instructions, reduce the resolution gap and its impact on simulated emergency drills, and can also improve the flexibility of simulated emergency drills and their coordination with real-time environments.

[0009] One advantage of the present application is that it provides an emergency decision-making simulation system based on behavior trees and multi-resolution deaggregation, wherein the emergency decision-making simulation system based on behavior trees and multi-resolution deaggregation makes full use of the modular and reusable characteristics of behavior trees, adds behavior tree nodes, and constructs emergency decision-making rules through flexible assembly and configuration of multiple groups of behavior tree nodes, aiming to realistically simulate the emergency drill process. When the entity state space increases, only a subtree needs to be added on the basis of the original behavior tree, and a single node of the behavior tree can be reused, which can flexibly expand the behavior tree structure and improve the reuse rate of the module.

[0010] One advantage of the present application is that it provides an emergency decision-making simulation system based on behavior trees and multi-resolution deaggregation, wherein the decision simulation system in the emergency decision-making simulation system based on behavior trees and multi-resolution deaggregation includes a decision rule editor and a decision scheduling engine, wherein the decision rule editor flexibly enters expert experience knowledge in a visual manner, can be decoupled from the software code, and avoids the "hard coding" mode of existing decision rules; the decision scheduling engine schedules and processes decision rules, and can be decoupled from the simulation system through a data subscription interface, and is not constrained by the version of the simulation system.

[0011] One advantage of the present application is that it provides an emergency decision-making simulation system based on behavior trees and multi-resolution deaggregation, wherein the emergency decision-making simulation system based on behavior trees and multi-resolution deaggregation utilizes a hierarchical concept to divide the emergency decision-making process into a mission layer, a task layer, and an action layer, each layer being responsible for a specific function, and converting a low-resolution emergency task into a hierarchical behavior tree that can drive a model, and deriving the decision result by combining hierarchical behavior trees, thereby deaggregating into a set of interrelated high-resolution model instructions, and designing situation processing nodes, blackboard decoration nodes, threat assessment nodes, and reference nodes in the behavior tree structure, which can improve the assembly and configuration flexibility of the behavior tree and improve the coordination between simulated emergency drills and real-time environments.

[0012] According to one aspect of the present application, an emergency decision simulation system based on behavior tree and multi-resolution deaggregation is provided, which includes: a decision simulation system and a simulation deduction system, wherein the decision simulation system and the simulation deduction system can be interactively connected to each other; the decision simulation system includes a decision rule editor and a decision scheduling engine, the decision rule editor is used to generate a decision rule script file, and the decision scheduling engine is communicatively connected to the decision rule editor, and is used to obtain simulation deduction data in real time and generate action instructions based on the decision rule script file of the decision rule editor and send the action instructions to the simulation deduction system;

[0013] The decision rule editor includes a mission layer, a task layer and an action layer which are communicatively connected in sequence; the mission layer is used to abstract the emergency handling intention into an emergency mission based on the event state in the emergency simulation; the task layer is used to decompose the emergency mission of the mission layer into detailed tasks; the action layer is used to configure actions based on the detailed tasks of the task layer; the mission layer includes a mission layer behavior tree; the task layer includes a task layer behavior tree; the action layer includes an action layer behavior tree; the mission layer behavior tree, the task layer behavior tree and the action layer behavior tree are linked through reference nodes;

[0014] Among them, the mission layer behavior tree uses the prevention and control reference node, the search and rescue reference node, the first aid reference node, and the transfer reference node as the lowest level node of the mission layer behavior tree and the root node of the task layer behavior tree, and is provided with a situation processing node and a blackboard decoration node.

[0015] In one embodiment of the emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation described in the present application, the situation processing node and the blackboard decoration node are arranged at the upper level of the lowest level node of the mission layer behavior tree, and the situation processing node is used to write the status information of the leaf node of the action layer behavior tree into the blackboard decoration node; the status information includes the simulation entity model status, action execution status and task target completion status of the simulation system.

[0016] In one embodiment of the emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation described in the present application, the task layer behavior tree includes a node for determining whether it is suitable for normal prevention and control and a node for determining whether it is suitable for emergency prevention and control, as well as an emergency prevention and control reference node; and a threat assessment node and a selection node at the same level as the threat assessment node are provided.

[0017] In one embodiment of the emergency decision-making simulation system based on behavior trees and multi-resolution deaggregation described in the present application, the threat assessment node is at the same level as the node for determining whether it is suitable for normal prevention and control, the node for determining whether it is suitable for emergency prevention and control, and the emergency prevention and control reference node. The threat assessment node is used to analyze the disaster threat level and spatial distance for the task layer to determine the prevention and control method to be adopted.

[0018] In one embodiment of the emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation described in the present application, the next-level nodes of at least one of the selection nodes at the same level as the threat assessment node include conventional prevention and control reference nodes and placeholder prevention and control reference nodes.

[0019] In one embodiment of the emergency decision-making simulation system based on behavior trees and multi-resolution deaggregation described in the present application, the next-level node of at least one of the selection nodes at the same level as the threat assessment node includes a node for determining whether there is a prevention and control target and a node for selecting a prevention and control target.

[0020] In one embodiment of the emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation described in the present application, the preset disaster threat levels include low, medium and high, and high; the preset spatial distances include long distance, medium and long distance, and close distance; when the disaster threat level is assessed as low and the spatial distance is assessed as long distance, the task layer adopts conventional prevention and control methods, and the action layer behavior tree determines the action configuration based on the conventional prevention and control reference node; when the disaster threat level is assessed as medium and high and / or the spatial distance is assessed as medium and long distance, the task layer adopts the placeholder prevention and control method, and the action layer behavior tree determines the action configuration based on the placeholder prevention and control reference node; when the disaster threat level is assessed as high and the spatial distance is assessed as close distance, the task layer adopts the emergency prevention and control method, and the action layer behavior tree determines the action configuration based on the emergency prevention and control reference node.

[0021] In one embodiment of the emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation described in the present application, the decision rule editor is used to enter expert experience knowledge in a visual manner and is suitable for decoupling from software code.

[0022] In one embodiment of the emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation described in the present application, the decision scheduling engine is suitable for being decoupled from the simulation system through a data subscription interface.

[0023] Further objectives and advantages of the present application will be fully reflected through understanding of the following description and drawings.

[0024] These and other objects, features and advantages of the present application are fully reflected in the following detailed description, drawings and claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or.

[0026] Figure 1 The figure illustrates a block diagram of an emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation according to an embodiment of the present application.

[0027] Figure 2 The figure shows a block diagram of the emergency implementation plan in emergency response implementation.

[0028] Figure 3Block diagram illustrating the resolution gap between emergency tasks and model instructions.

[0029] Figure 4 The figure illustrates a schematic diagram of a modeling block diagram of an emergency decision-making simulation system based on behavior trees and multi-resolution deaggregation according to an embodiment of the present application.

[0030] Figure 5 The figure illustrates a block diagram of a mission layer behavior tree of an emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation according to an embodiment of the present application.

[0031] Figure 6 A block diagram of the local structure of a task-layer behavior tree of an emergency decision-making simulation system based on behavior trees and multi-resolution deaggregation according to an embodiment of the present application is illustrated.

[0032] Figure 7 A block diagram of the local structure of a behavior tree at the behavior layer of an emergency decision-making simulation system based on behavior trees and multi-resolution deaggregation according to an embodiment of the present application is illustrated. DETAILED DESCRIPTION

[0033] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0034] It is understood that the term "one" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element can be one, while in another embodiment, the number of the element can be multiple, and the term "one" cannot be understood as a limitation on the number. "Multiple" means greater than or equal to two.

[0035] Although ordinals such as "first," "second," and the like will be used to describe various components, those components are not limited herein. The term is used only to distinguish one component from another. For example, a first component may be referred to as a second component, and likewise, a second component may be referred to as a first component without departing from the teachings of the present application. The term "and / or" as used herein includes any and all combinations of one or more associated listed items.

[0036] The terms used herein are only used for the purpose of describing various embodiments and are not intended to be limiting. As used herein, the singular form is intended to include the plural form, unless the context clearly indicates an exception. It will also be understood that the terms "including" and / or "having" when used in this specification specify the presence of the described features, numbers, operations, components, elements, or combinations thereof, without excluding the presence or addition of one or more other features, numbers, operations, components, elements, or combinations thereof.

[0037] As mentioned above, the current research process of simulation emergency decision-making deduction modeling faces the following difficulties: (1) Traditional simulation decision-making research is mainly carried out from two aspects: internal and external autonomous decision-making. Internal decision-making relies on the decision logic solidified in the model to realize the understanding of the emergency situation, and external decision-making realizes the simulation decision-making process by entering expert knowledge and experience. However, the development of internal decision-making is difficult, and it is difficult to combine expert knowledge and experience in the model. Although external decision-making uses expert experience, it is often based on finite state machines to implement rule decisions. When the scale of the state space increases, it is difficult for finite state machines to maintain complex rules and condition description spaces. Behavior trees are an improvement of finite state machines, supporting the addition of single behavior nodes or behavior subtrees. When faced with complex emergency response scenarios, the constructed behavior trees have problems such as poor readability, strong coupling between different modules, and fixed structure that is difficult to adapt to situation changes. (2) The resolution of the model in the current mainstream simulation system is generally at the platform level. However, in the emergency response tasks issued by the emergency commander, the execution unit scale is large, and the corresponding simulation system entity model resolution is low, resulting in a resolution gap between the emergency response plan issued by the high-level emergency commander and the simulation system model instructions. In addition, emergency response plans have multiple levels of mission, task, and action, which makes the plan have multiple resolutions, increasing the resolution level of this difference. (3) The variability of the emergency response environment, the diversity of emergencies, the uncertainty of the strategies for the objects of emergencies, and the existence of multi-level decision-making processes require the emergency response decision-making model to make highly complex decisions in real-time situations. Therefore, it is very cumbersome and difficult to construct a mapping from the current environment to purpose-oriented actions.

[0038] In view of the problems faced by traditional rule-based decision-making methods, such as complex and difficult description of decision rules and conditions when the scale of state space increases, the resolution gap between high-level emergency response tasks and simulation model instructions, and multi-level emergency decision-making processes, this application proposes an emergency decision-making deduction modeling scheme based on behavior trees and multi-resolution decomposition. First, the modular combination, scalability, and interpretability of behavior trees are fully utilized. On the basis of the original behavior tree nodes, situation processing nodes, blackboard nodes, threat assessment nodes, and reference nodes are designed to edit intelligent confrontation decision rules; then, using the idea of ​​​​layering, the emergency response decision-making process is divided into mission layer, task layer, and action layer, and the reference nodes are used to link the three layers to form a composite behavior tree structure, thereby decomposing low-resolution emergency tasks into high-resolution model instructions; finally, based on the emergency decision-making simulation system with behavior trees and multi-resolution decomposition, the interaction between the decision simulation system and the simulation deduction system is completed.

[0039] like Figures 1 to 7 As shown, the emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation according to the embodiment of the present application is explained. Figure 1 As shown, according to an embodiment of the present application, the emergency decision simulation system based on behavior tree and multi-resolution deaggregation includes a decision simulation system and a simulation deduction system, wherein the decision simulation system and the simulation deduction system can be interactively connected to each other; the decision simulation system includes a decision rule editor and a decision scheduling engine, the decision rule editor is used to generate a decision rule script file, and the decision scheduling engine can be communicatively connected to the decision rule editor, and is used to obtain simulation deduction data in real time and generate action instructions based on the decision rule script file of the decision rule editor and send the action instructions to the simulation deduction system.

[0040] Below, the construction principle and construction process of the emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation of the present application are described in more detail.

[0041] Specifically, the emergency decision simulation system based on behavior tree and multi-resolution decomposition of the present application applies the idea of ​​"simulation deduction" to emergency disposal. Emergency disposal implementation includes emergency task planning and action instruction execution, which have fractal characteristics in process structure. Emergency response units in emergency disposal are faced with the problem of choosing the next action. Before taking new actions, they will always make decisions based on factors such as tasks, plans, and situation information. The decision-making process of the emergency response unit consists of multiple "perception-decision-action" behavior sequences. Each emergency response unit uses various sensors to obtain information about the surrounding environment, and the emergency disposal commander makes decisions based on this, and then takes various actions (including search and rescue, treatment, etc.). The "decision" of the emergency disposal commander determines what actions the emergency response unit takes and what information will be perceived. The "decision-making" link directly determines "why this behavior should be taken" and "what actions can achieve the emergency disposal goal". The information structure process of the emergency disposal commander's cognition of the implementation of emergency response is gradual, so the decision-making process of emergency disposal implementation is hierarchical.

[0042] According to the organizational structure and command and control mode of emergency implementation planning, emergency implementation planning includes three levels: commander planning, team leader planning and emergency response unit planning, such as Figure 2 As shown in the figure, the schematic diagram of the emergency implementation planning framework in emergency response implementation shows the hierarchy of the emergency implementation planning.

[0043] The commander of the headquarters determines the emergency area, emergency mission, etc. according to the overall situation of the environment and emergency forces, selects the appropriate disposal department and assigns tasks to complete the department task planning. The team leader of the disposal department assigns emergency goals and coordination plans to the emergency response units within the department based on the received disposal tasks, taking into account the actual environmental situation and emergency response capabilities, and completes the department disposal plan. The individual emergency response units independently complete emergency actions such as search and rescue, first aid, and evacuation according to the assigned goals. The disposal situation of the emergency response unit will be fed back to the emergency commanders at different levels step by step, and the emergency commanders at all levels will plan emergency actions according to the emergency situation to achieve the mission. Therefore, in view of the hierarchical nature of emergency planning, it is necessary to propose a corresponding emergency decision-making method to realize emergency intelligent planning.

[0044] The emergency decision model is the core of the decision simulation system. The resolution reflects the precision of the emergency decision model in describing objective things such as the emergency environment and emergency response units. Combining different resolution models of the same thing for simulation can effectively improve the realism of the simulation results. This application constructs an emergency decision model based on behavior trees and multi-resolution disaggregation.

[0045] Human-in-the-loop simulation is a typical application mode of emergency decision-making simulation system. The commander of the headquarters controls the physical model of the simulation system to perform emergency actions by issuing instructions. However, in the emergency tasks issued by the high-level commander (for example, the emergency commander of the headquarters), the execution unit is large in scale, and the corresponding physical model resolution of the simulation system is low. This will cause a resolution gap problem between the emergency tasks issued by the high-level commander and the model instructions of the simulation system, such as Figure 3 As shown in the figure. When high-level commanders use simulation and deduction systems for confrontational deductions, they do not command subordinate departments at the same level, but extend downward to entities such as emergency resources, and bear a lot of additional command work. Therefore, it is necessary to decompose low-resolution emergency tasks into high-resolution model instructions with the model instructions of the simulation and deduction system as the center, so as to improve the human-computer interaction experience between high-level commanders and the simulation and deduction system.

[0046] Constructing a mapping from the current emergency environment to actions oriented towards emergency goals is cumbersome and difficult. This application uses a hierarchical approach to divide the emergency decision-making simulation system based on behavior trees and multi-resolution deaggregation into several levels. Each level is responsible for a specific function, and the multi-level decision-making process can make the emergency implementation plan easier to manage and maintain. After analyzing the emergency implementation plan, the commander's plan, the team leader's plan, and the emergency response unit plan correspond to the three levels of emergency response mission, task, and action, respectively. Accordingly, the emergency decision-making simulation system based on behavior trees and multi-resolution deaggregation can be modeled according to the four levels of mission, task, action, and model, such as Figure 4 .

[0047] The mission level is a general term for the predetermined goals, determinations, and plans of emergency actions. It is used to express the emergency tasks assigned to a certain level of commander, and is usually a reflection of the intentions of the superiors. Emergency missions are macro and general, and cannot be understood and executed by grassroots emergency response units. They need to be broken down and explained in detail by a certain level of commander.

[0048] The task layer is an action plan drawn up according to the emergency mission. The task is the phased behavior that can be divided according to the actual emergency situation during the mission decomposition process. The task layer analyzes and evaluates the environmental situation, creates new tasks or updates existing tasks, and dispatches existing available emergency resources to perform tasks.

[0049] The action layer refers to the emergency actions taken according to the actual situation of the task and environment. It is a relatively independent and indecomposable action behavior, which provides specific action support for the execution of the task and issues action instructions to the emergency simulation system.

[0050] The model layer refers to the high-resolution simulation entity model provided by the emergency simulation platform, and the action layer controls the simulation entity model through action instructions.

[0051] Accordingly, the decision rule editor includes a mission layer, a task layer and an action layer that are communicatively connected in sequence. The present application decomposes low-resolution emergency tasks into high-resolution model instructions through system stratification, thereby reducing the resolution gap and its impact on simulated emergency drills.

[0052] This application makes full use of the behavior tree with advantages such as modular combination, scalability, and interpretability to construct a mission layer, task layer, and action layer with high flexibility, so as to improve the flexibility of simulated emergency drills and their coordination with the real-time environment.

[0053] Behavior Tree (BT) is a hierarchical node tree used to control the decision-making process of simulation entities. Leaf nodes are used to control the behavior of simulation entity models, and branch nodes represent conditions or control structures. The type of parent node determines the execution logic relationship between child nodes and controls the execution status of child nodes. Behavior Tree is a directed tree consisting of nodes and directed edges, defined as a binary BT=<V,E> , where: V = A∪C∪N∪τ, Among them, A represents action node; C represents condition node; N represents control flow node; τ represents root node; V represents finite node set, including action node, condition node, control flow node and root node. Action node and condition node are leaf nodes of behavior tree, which are used to express the specific behavior instructions of simulation entity model. Logic node is non-leaf node, which determines how simulation entity model moves from root node to leaf node along a certain path according to different situations. E is finite edge set, Among them, vi Yes j The parent node of .

[0054] The nodes of the behavior tree include combination nodes, decorator nodes, and leaf nodes. Combination nodes, also known as "control nodes", are non-leaf nodes that control the direction of the traversal path and manage the execution of child nodes. There are many types of them, including selection nodes, sequence nodes, parallel nodes, and random nodes. Selection nodes will return the result of the current child node when any child node returns success, and end the selection operation, and will no longer continue to execute subsequent child nodes at the same level. Selection nodes will only return failure when all child nodes return failure. Sequence nodes will visit all child nodes in sequence. If all child nodes return success, the sequence node returns success; if any child node returns failure during the traversal, the subsequent child nodes will no longer be executed. Parallel nodes will execute multiple child nodes at the same time. Decorator nodes are mainly used to modify the result returned by child nodes, terminate child nodes, or repeat child nodes and other additional conditions. Common decorator nodes include inversion nodes, success nodes, and repeat nodes. Leaf nodes are the end nodes on each branch, including behavior nodes and condition nodes. Its function is to make the simulation entity do specific things. It realizes complex and intelligent behavior logic by collaborating with combination nodes and modification nodes.

[0055] During the simulation cycle, the behavior tree will traverse from the root node to the leaf nodes of each subtree in a top-down and left-to-right order. The nodes that have been executed will not affect the direction of the execution branch, and there will be no mutual interference between the various behavior subtrees, which can improve the modularity of simulation decision modeling.

[0056] In emergency response simulation, behavior trees can abstract the decision-making process into behavior nodes, which can be combined with conditional nodes according to emergency plans. Complex decision-making processes can also be combined into hierarchical behavior trees to deduce decision results. Therefore, behavior trees can be used to clearly represent and organize complex decision logic and behavior processes.

[0057] An emergency task is a description of emergency activities. The tasks issued by superiors or senior management are often macro-summary tasks. Emergency simulation must break down the macro-tasks into specific details until there is a simulation entity that can be executed. This is the emergency task decomposition process.

[0058] The low-resolution emergency tasks issued by the high-level emergency commander are decomposed into high-resolution model instructions, and the simulation model can execute the instructions to generate the emergency actions corresponding to the emergency tasks. Traditional emergency task decomposition is mainly through the template method, but this method cannot control the simulation model to execute the corresponding instructions. The behavior tree supports the addition of a single behavior node or behavior subtree, and has modularity and reusability. It can aggregate high-resolution model instructions into low-resolution emergency tasks through module combination, and can obtain the status information of the leaf node execution action in real time, and can monitor the execution status of the simulation entity model of the simulation deduction system in real time. The hierarchical behavior tree structure can be used to realize the decomposition of low-resolution emergency tasks into high-resolution model instructions with the model instructions of the simulation deduction system as the center, so as to solve the gap problem between low-resolution emergency tasks and high-resolution model instructions.

[0059] The low-resolution emergency tasks are decomposed into high-resolution model instructions. These instructions break down the tasks into specific simulation model execution steps, forming a specific action plan that describes which emergency response unit is assigned to do what at what time and where. To this end, it is also necessary to expand the local state monitoring in the traditional behavior tree to global state monitoring and increase the available emergency resource scheduling decision-making capabilities.

[0060] In actual emergency drills, emergency handling is a very complex hierarchical decision-making process. The top-level behavior of the emergency response unit is based on the emergency task assigned by the superior, which defines the activity area and emergency principles of the emergency response unit from a macro perspective. The emergency response unit has main action plans and backup action plans under different circumstances for each emergency task. After determining the action plan, the emergency commander determines the main behavior of the emergency response unit based on the actual situation. Then determine more detailed, more instructive and more operational actions. Finally, the emergency simulation process is implemented in the specific actions of the emergency response unit, that is, a series of rescue and transfer actions performed at any time, in any environment and in any event. Using the purpose-oriented emergency planning design and task hierarchical method, the mission layer, task layer, and action layer of the proposed overall framework of emergency decision modeling are designed into complete behavior trees respectively. The layers are linked by referencing nodes to build an efficient organizational structure and provide reusable decision-making strategies.

[0061] Correspondingly, the mission layer includes a mission layer behavior tree; the task layer includes a task layer behavior tree; the action layer includes an action layer behavior tree; the mission layer behavior tree, the task layer behavior tree and the action layer behavior tree are linked through reference nodes, which can improve the assembly configuration flexibility of the behavior tree.

[0062] The emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation in this application makes full use of the modular and reusable characteristics of behavior tree, adds new behavior tree nodes, and constructs emergency decision-making rules through flexible assembly and configuration of multiple groups of behavior tree nodes, aiming to realistically simulate the emergency drill process. When the entity state space increases, only a subtree needs to be added on the basis of the original behavior tree, and a single node of the behavior tree can be reused, which can flexibly expand the behavior tree structure and improve the reuse rate of the module.

[0063] In the mission layer, the emergency response intention is abstracted into prevention and control, search and rescue, first aid, transfer, etc. according to the deduction score, triggering events and other states in the emergency simulation. Prevention and control refers to preventing the spread or aggravation of disasters, search and rescue refers to quickly locating trapped people or dangerous sources, first aid refers to giving priority to medical care and life support for the injured, and transfer refers to quickly and safely evacuating people from dangerous areas.

[0064] Accordingly, the mission layer is used to abstract the emergency handling intention into an emergency mission based on the event status in the emergency simulation. Figure 5 As shown, the mission layer behavior tree uses the prevention and control reference node, the search and rescue reference node, the first aid reference node, and the transfer reference node as the lowest level node of the mission layer behavior tree and the root node of the task layer behavior tree; and newly adds a situation processing node and a blackboard decoration node. Specifically, the situation processing node and the blackboard decoration node are set at the upper level of the lowest level node of the mission layer behavior tree, that is, the upper level of the prevention and control reference node, the search and rescue reference node, the first aid reference node, and the transfer reference node.

[0065] The situation processing node obtains the situation of the simulation process through the simulation system, and writes the model state, instruction execution state and task target completion state involved in the leaf node of the action layer behavior tree in the situation into the blackboard decoration node.

[0066] Correspondingly, the situation processing node is used to write the status information of the leaf node of the action layer behavior tree into the blackboard decoration node, which can improve the coordination between the simulated emergency drill and the real-time environment; the status information includes the simulation entity model status, action execution status and task target completion status of the simulation deduction system.

[0067] The blackboard decoration node is a shared memory mechanism between different nodes of the behavior tree, which meets the communication requirements of the behavior subtree.

[0068] The task layer is used to decompose the emergency mission based on the mission layer into detailed tasks. The detailed tasks are different according to the different action purposes of the emergency response unit. In an example of the present application, the detailed tasks of the task layer include: environmental reconnaissance, disaster prevention and control, and logistical support. Environmental reconnaissance refers to the assessment of the environmental conditions of disaster-stricken or dangerous areas to provide safety guarantees for subsequent rescue operations; disaster prevention and control refers to controlling the source of disasters to prevent the spread or intensification of disasters and reduce the occurrence of secondary disasters; logistical support refers to providing materials, equipment and technical support to rescue teams and disaster-stricken personnel. The behavior tree under the prevention and control mission in the task layer is as follows: Figure 6 As shown. The behavior tree under the prevention and control mission in the task layer takes the prevention and control reference node of the mission layer behavior tree as the root node, and different detailed tasks are linked by combination nodes. The same level of the reference node is the decision information node, which provides decision information for the execution of different tasks.

[0069] The behavior tree structure under the rescue, first aid, and transfer missions in the task layer is similar to the behavior tree structure under the prevention and control mission. Accordingly, the behavior tree under the rescue mission in the task layer uses the rescue reference node of the mission layer behavior tree as the root node; the behavior tree under the first aid mission in the task layer uses the first aid reference node of the mission layer behavior tree as the root node; the behavior tree under the transfer mission in the task layer uses the transfer reference node of the mission layer behavior tree as the root node.

[0070] It is worth mentioning that in the embodiment of the present application, a threat assessment node and a selection node at the same level as the threat assessment node are provided in the task layer behavior tree to consider the behavioral characteristics of the simulation entity model and comprehensively utilize situation analysis to determine the emergency response method.

[0071] Specifically, the task layer behavior tree includes a node for determining whether it is suitable for normal prevention and control, a node for determining whether it is suitable for emergency prevention and control, and the emergency prevention and control reference node. The threat assessment node is at the same level as the node for determining whether it is suitable for normal prevention and control, the node for determining whether it is suitable for emergency prevention and control, and the emergency prevention and control reference node.

[0072] The threat assessment node is used to analyze the disaster threat level and spatial distance so that the task layer can determine the prevention and control methods to be adopted, such as conventional prevention and control, positional prevention and control, or emergency prevention and control, which facilitates the task layer to solve and prepare various prevention and control problems, and can improve the coordination between simulated emergency drills and real-time environments.

[0073] Correspondingly, the next-level nodes of at least one of the selection nodes at the same level as the threat assessment node include conventional prevention and control reference nodes and placeholder prevention and control reference nodes.

[0074] Specifically, the preset disaster threat levels include low, medium and high, and high; the preset spatial distances include long distance, medium and long distance, and short distance. The low, medium and high levels of the disaster threat level can be determined according to the preset task layer behavior tree including nodes for determining whether it is suitable for normal prevention and control and nodes for determining whether it is suitable for emergency prevention and control, as well as the conditions of the emergency prevention and control reference node. For example, in the implementation of earthquake emergency response, the disaster threat level of earthquakes with a magnitude less than 4.5 can be set to a low level, the disaster threat level of earthquakes with a magnitude greater than or equal to 4.5 and less than 7 can be set to a medium and high level, and the disaster threat level of earthquakes with a magnitude greater than or equal to 7 can be set to a high level.

[0075] The long distance, medium-long distance and short distance of spatial distance can be determined according to preset conditions. For example, in the implementation of earthquake emergency response, the distance greater than 1,000 kilometers can be set as the long distance, the distance less than 1,000 kilometers and greater than or equal to 500 kilometers can be set as the medium-long distance, and the distance less than 500 kilometers can be set as the short distance.

[0076] It should be understandable that the disaster threat level and spatial distance can be set according to the actual scenario. For example, the disaster threat level can be divided into low, medium-high, high and very high, and the spatial distance can be divided into long distance, medium-long distance, relatively close distance and close distance.

[0077] The prevention and control methods corresponding to different disaster threat levels and spatial distances can be determined according to actual needs. In an example of the present application, when the disaster threat level is assessed as low and the spatial distance is assessed as long, the task layer adopts a conventional prevention and control method, and the action layer behavior tree determines the action configuration based on the conventional prevention and control reference node; when the disaster threat level is assessed as medium-high and / or the spatial distance is assessed as medium-long, the task layer adopts a place-based prevention and control method, deploys rescue forces and emergency equipment to key locations in advance, and prepares to initiate prevention and control measures at the best time. The action layer behavior tree determines the action configuration based on the place-based prevention and control reference node. In the process of determining whether to adopt the place-based prevention and control method, the threat duration and threat trend can be further introduced for a more accurate judgment. For example, when the disaster threat level is assessed as medium-high and / or the spatial distance is assessed as medium-long and the threat may be upgraded or expanded in the future, the task layer adopts a place-based prevention and control method, and the action layer behavior tree determines the action configuration based on the place-based prevention and control reference node. When the disaster threat level is assessed as high and the spatial distance is assessed as close, the task layer adopts emergency prevention and control methods, such as blocking and isolating the disaster-stricken area to prevent the disaster from expanding, evacuating and evacuating personnel, and avoiding casualties. The action layer behavior tree determines the action configuration based on the emergency prevention and control reference node.

[0078] Whether there is a control target and the selection of a control target are the premise and basis of control. Under the condition of determining whether there is a control target or selecting a control target, the selection and determination of the control target can be realized according to the intention of the control command. Accordingly, the next-level node of at least one of the selection nodes at the same level as the threat assessment node includes a node for determining whether there is a control target and a node for selecting a control target.

[0079] In emergency simulation, action refers to the specific action performed by the simulation entity in accordance with the instructions. Accordingly, the action layer is used to configure actions based on the detailed tasks of the task layer. When the action layer behavior tree determines the action configuration based on the placeholder prevention and control reference node, when the simulation entity model adopts placeholder prevention and control, it needs to perform actions such as path planning, equipment deployment, establishment of isolation barriers, and area monitoring. Tasks are often composed of a combination of transfer, alert, or complex strategic actions. The behavior tree of the action layer behavior tree under the placeholder prevention and control task is as follows: Figure 7 As shown in the figure, actions such as path planning and equipment deployment are arranged in a certain order through combined nodes. The order of action nodes reflects which behavior the simulation entity model should perform in different situations to complete the task.

[0080] In the emergency decision-making model based on behavior tree and multi-resolution deaggregation, reference nodes are used to link behavior subtrees at different levels, and combination nodes are used to link behavior subtrees at the same level to build a multi-level behavior tree structure. The hierarchical behavior tree structure decouples complex emergency decisions into behavior subtrees with low correlation between missions, tasks, and actions. Subtrees at each level can be maintained and adjusted separately, increasing the flexibility of the decision-making system. Behavior subtrees of the same type can be freely combined through reference nodes, and different types of behavior subtrees can call other types of behavior subtrees, improving the reusability of the decision-making system.

[0081] The constructed emergency decision model based on behavior tree and multi-resolution deaggregation forms the core part of the decision simulation system of the present application, and forms the decision rule editor of the decision simulation system. The decision rule editor provides a graphical drag-and-drop method to assist emergency personnel in editing emergency behavior trees and generating decision rule script files. The decision rule editor is used to input expert experience knowledge in a visual way, and is suitable for decoupling from software code to avoid the "hard coding" mode of existing decision rules.

[0082] Emergency decision-making simulation is a dynamic conditional judgment, a process that follows behavioral rules and is constantly adjusted. It is necessary to obtain real-time simulation data from the simulation system, determine the execution condition nodes, determine the behavioral actions to be executed, generate behavioral action instructions, and issue them to the simulation entity model in the simulation system to trigger the simulation entity model to execute according to the action instructions. This application realizes the interaction between the decision simulation system and the simulation system through a decision scheduling engine.

[0083] The decision scheduling engine mainly completes the creation and operation scheduling management of behavior tree objects, loads and parses decision rule script files, and receives the simulation data generated by the simulation deduction system in real time. The emergency decision simulation system based on behavior tree and multi-resolution deaggregation starts traversing from the root of the behavior tree every time the simulation clock advances, and checks whether each node is activated in order from top to bottom and from left to right until the current activated node refreshes the behavior tree.

[0084] The decision scheduling engine interacts with the simulation and deduction system. On the one hand, the simulation and deduction data are acquired in real time, wherein the simulation and deduction data include the simulation entity model status data such as the target longitude and latitude, speed, and direction of the simulation entity model, the simulation entity model detection data such as the detection target type, the detection target longitude and latitude, the detection target speed, the detection target direction, and the action status data such as the action execution result; on the other hand, the action instructions generated by the decision scheduling engine are sent to the simulation and deduction system to control the simulation entity model to perform emergency actions.

[0085] The decision scheduling engine is suitable for being decoupled from the simulation system through a data subscription interface and is not constrained by the version of the simulation system.

[0086] In summary, according to the embodiment of the present application, an emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation is explained, and the emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation can deaggregate low-resolution emergency tasks into high-resolution model instructions, reduce the resolution gap and its impact on simulated emergency drills, and can also improve the flexibility of simulated emergency drills and their coordination with real-time environments.

[0087] The present application and its implementation methods are described above, and such description is not restrictive. The drawings show only one implementation method of the present application, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it and design structural methods and embodiments similar to the technical solution without creative design without departing from the inventive purpose of the present application, they should all fall within the protection scope of the present application.

Claims

1. An emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation, characterized in that: The invention comprises a decision simulation system and a simulation deduction system, wherein the decision simulation system and the simulation deduction system can be interactively connected to each other; the decision simulation system comprises a decision rule editor and a decision scheduling engine, wherein the decision rule editor is used to generate a decision rule script file, and the decision scheduling engine can be communicatively connected to the decision rule editor, and is used to obtain simulation deduction data in real time and generate action instructions based on the decision rule script file of the decision rule editor, and send the action instructions to the simulation deduction system; wherein the decision rule editor comprises a mission layer, a task layer and an action layer which are communicatively connected in sequence; the mission layer is used to generate a decision rule script file based on the event status in the emergency simulation; The emergency response intention is abstracted into an emergency mission; the task layer is used to decompose the emergency mission based on the mission layer into detailed tasks; the action layer is used to configure actions based on the detailed tasks of the task layer; the mission layer includes a mission layer behavior tree; the task layer includes a task layer behavior tree; the action layer includes an action layer behavior tree; the mission layer behavior tree, the task layer behavior tree and the action layer behavior tree are linked through reference nodes; wherein the mission layer behavior tree uses a prevention and control reference node, a search and rescue reference node, a first aid reference node, and a transfer reference node as the lowest level node of the mission layer behavior tree and the root node of the task layer behavior tree, and is provided with a situation processing node and a blackboard decoration node.

2. The emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation according to claim 1 is characterized in that: The situation processing node and the blackboard decoration node are arranged at the upper level of the lowest level node of the mission layer behavior tree. The situation processing node is used to write the status information of the leaf node of the action layer behavior tree into the blackboard decoration node; the status information includes the simulation entity model status, action execution status and task target completion status of the simulation deduction system.

3. The emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation according to claim 2 is characterized in that: The task layer behavior tree includes a node for determining whether it is suitable for normal prevention and control, a node for determining whether it is suitable for emergency prevention and control, and an emergency prevention and control reference node; and is provided with a threat assessment node and a selection node at the same level as the threat assessment node.

4. The emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation according to claim 3 is characterized in that: The threat assessment node is at the same level as the node used to determine whether it is suitable for normal prevention and control, the node used to determine whether it is suitable for emergency prevention and control, and the emergency prevention and control reference node. The threat assessment node is used to analyze the disaster threat level and spatial distance for the task layer to determine the prevention and control method to be adopted.

5. The emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation according to claim 4 is characterized in that: The next-level nodes of at least one of the selection nodes at the same level as the threat assessment node include conventional prevention and control reference nodes and placeholder prevention and control reference nodes.

6. The emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation according to claim 5 is characterized in that: The next-level nodes of at least one of the selection nodes at the same level as the threat assessment node include a node for determining whether there is a prevention and control target and a node for selecting a prevention and control target.

7. The emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation according to claim 5 is characterized in that: The preset disaster threat levels include low, medium and high, and high; the preset spatial distances include long distance, medium and long distance, and short distance; when the disaster threat level is assessed as low and the spatial distance is assessed as long distance, the task layer adopts conventional prevention and control methods, and the action layer behavior tree determines the action configuration based on the conventional prevention and control reference node; when the disaster threat level is assessed as medium and high and / or the spatial distance is assessed as medium and long distance, the task layer adopts placeholder prevention and control methods, and the action layer behavior tree determines the action configuration based on the placeholder prevention and control reference node; when the disaster threat level is assessed as high and the spatial distance is assessed as short distance, the task layer adopts emergency prevention and control methods, and the action layer behavior tree determines the action configuration based on the emergency prevention and control reference node.

8. The emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation according to claim 1 is characterized in that: The decision rule editor is used to input expert experience knowledge in a visual manner and is suitable for being decoupled from software codes.

9. The emergency decision-making simulation system based on behavior tree and multi-resolution deaggregation according to claim 1 is characterized in that: The decision scheduling engine is suitable for being decoupled from the simulation system through a data subscription interface.

Images