Joint action simulation method adopting deterministic conditional Monte Carlo simulation scheme and solver system

Abstract

The invention provides a joint action simulation method adopting a deterministic condition Monte Carlo simulation scheme and a solver system, and the method comprises the steps: Monte Carlo simulation: carrying out the traversal calculation of a key element combination for an action target to be simulated through employing the idea of deterministic condition Monte Carlo simulation, inputting the combination of all the key parameters into a simulation entity for calculation by utilizing the high-speed calculation capability of a computer system, finally automatically obtaining n results, and then screening m optimal results from the n results; a virtual clock metronome step: designing and coordinating a virtual clock metronome of each simulation module, wherein the virtual clock metronome is used for simulating tasks required to be completed in the real world by each module within the same time interval; and a perception interaction step: carrying out simulation entity perception layer interaction by adopting an artificial intelligence mode. The method has the advantages that the deterministic condition Monte Carlo simulation scheme is used for tactical-level action simulation, and wide application prospects and value are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of tactical simulation, and in particular to a joint action simulation method and a solver system using a deterministic conditional Monte Carlo simulation scheme. Background Art

[0002] 1. Historical Development of Tactical-Level Computer Action Simulation

[0003] The earliest forms of combat modeling and simulation can be traced back to the origins of human civilization and warfare. Before the advent of computer technology, early combat simulations used wargames to teach potential rulers and nobles about combat concepts and principles. In 1664, Germans developed specialized wargames to depict specific combat scenarios rather than abstract strategic ones. To this day, various wargames are still used to depict the effective actions of military units. In 1811, Prussian war advisor Reisswitz invented the first true wargame, featuring a tabletop terrain model that served as a miniature battlefield. Rules and diagrams governed the outcomes of actions in the game. To account for the random nature of combat outcomes, dice rolling was introduced as a random factor affecting the outcome of each action. The Kriegsspiel game, invented in 1837, allowed for discussion of the timing of combat actions and analysis of the outcomes of conflicts by depicting the unfolding battlefield. Thus, it incorporated the influence of space, time, and terrain on action. From then on, until the end of World War II, war games became, to some extent, the most effective means of "simulating" combat operations.

[0004] Computer wargaming systems began developing slowly in the 1960s and 1970s, but accelerated in the 1980s, ultimately surpassing traditional tabletop wargaming and replacing most military and civilian wargaming systems by the 1990s. The concept of modeling then emerged. First, aircraft modeling and simulation followed, followed by vehicle modeling and simulation, naval vessel operational simulation, and air defense weapon simulation. Operational simulation testing plays a crucial role in weapon and equipment development, enabling weapons and equipment developers to draw lessons from simulation and apply them to their designs. With the development of computer networks, distributed simulation system networks emerged, followed by the semi-automated force-based simulation system SIMNET. The Aggregation and Simulation Protocol (ALSP) was subsequently released, extending the advantages of distributed simulation to swarm force training. Finally, in the mid-1990s, the High Level Architecture (HLA) emerged to address the evolving needs of multi-service joint simulation and establish corresponding interface standards. This standard, with its complex architecture, requires all parties involved in joint operations to develop to a unified standard. It also became an open standard managed by the Simulation Interoperability Standards Organization (SISO).

[0005] The algorithms for action simulation calculations have also evolved to a certain extent, and mathematical models provide a theoretical basis for these calculation methods. These models can be very effective and are divided into two major types: (1) focusing on large-scale conflicts involving the use of a large number of weapons and equipment; (2) newer agent-based models. Early military planning decisions often used the Lanchester model, which analyzed the situation of air combat in World War I and developed a theory to explain why the concentration of forces would have an advantage in these complex military operations. This theory was later summarized into a law, which is to use a low-resolution aggregation model to determine the battle results of two forces fighting. This includes linear rate, square rate and additional parameters. The original Lanchester model did not include random behavior, and later based on practical applications, this defect was compensated and a random Lanchester model was produced. Later, due to the shortcomings of the Lanchester model, game theory was applied to action simulation. The action simulation system based on the agent model uses more modeling methods for complex systems and introduces a series of theoretical methods such as nonlinear dynamics, random dynamics, complexity theory, artificial life, evolution and genetic algorithms, cellular automata, neural networks, etc. to conduct research on the basic process of action.

[0006] With the continuous development of computer technology, more and more mathematical models, physical theoretical models and other algorithms are being applied to the modeling and simulation of joint actions.

[0007] It is worth noting that although joint action simulation originated from the military, in today's highly developed digital society, these technologies have been widely used in the civilian field and have broad application prospects and high application value.

[0008] 2. Architecture Types of Joint Action Simulation

[0009] Translating theory into practical engineering products requires first designing the system architecture. In the field of computer information systems, system architecture represents the correspondence between a system's functionality and its hardware and software elements, its software architecture and hardware architecture, and the interactions between people and these elements.

[0010] From another perspective, the system architecture is a comprehensive model and a systematic modeling method. Regarding joint action simulation, there are the following technical system architectures that are actually used. (1) Centralized architecture: The centralized architecture is a design pattern that was widely adopted before the emergence of distributed architecture. The centralized architecture is relatively simple. In the early stages of the development of action simulation training systems, combat simulation systems generally adopted a centralized architecture. In this architecture, the corresponding functions such as control, calculation or data are concentrated on a specific component or host. Systems using this type of architecture are more suitable for simulating equipment platforms or systems, such as ballistic simulation, aircraft combat effectiveness evaluation, etc. It is generally used for equipment development research and simulation.

[0011] (2) Distributed architecture: Centralized simulation systems cannot adapt to multi-platform tactical training, let alone the needs of multi-service synthesis and large-scale joint action training. The need to use cluster simulation methods for large-scale group collaborative training and research has prompted the emergence of distributed architecture. Of course, this also benefits from the development of computer network technology. "Distribution" means that simulation applications are distributed in different geographical locations and connected through a common communication infrastructure. "Interaction" is divided into two categories: the first is the data interaction required between multiple applications of different types, which is used to simulate the interaction in the actual action process. The interaction between systems includes the interaction between simulation entities (such as tanks, aircraft, etc.), such as engagement, avoidance, search, etc.; it also includes the mutual influence between simulation entities and the environment (atmosphere, ocean, terrain, outer space, etc.). In the traditional model, the interaction between systems is achieved by software exchanging data in a standard format according to a certain protocol. The second type of interaction is the interaction between people and simulation systems, that is, the interaction between people and systems. "Simulation" is to use computer programs and data to simulate various activities and events in real data as realistically as possible. In combat simulation, it mainly refers to combat operations. Distributed simulation systems are the mainstream architecture for joint operations simulation. Prominent examples include: High-Level Architecture (HLA); Model-Driven Architecture (MDA); Web-based Architecture (eXtensible Modeling and Simulation Framework, XMSF); and the Multi-Agent Based Modeling and Simulation (MABMS) modeling and simulation framework. It's worth noting that HLA is based on distributed object model information theory from the 1990s; MDA uses the Unified Modeling Language (UML) from the early 21st century to define models and promote conversion services, standards, and methods between different models; XMSF is based on more recent XML and Web Services technologies; and agent-based simulation frameworks are more rooted in intelligent information environments. It can be seen that the development of joint operations simulation technology has been incrementally evolving alongside advancements in information technology. Whenever a new information technology becomes popular, it is often applied to improve joint operations simulation technology.

[0012] 3. The driving effect of new technological development on this invention

[0013] Most of the currently used tactical-level joint (combat) action simulation systems originated in the mid-1990s. With the rapid development of information technology, various new information technologies have also been applied to action simulation systems. It is precisely the development of these new technologies that has made capabilities and solutions that were difficult to achieve in traditional simulation systems feasible. In other words, it is precisely this technological breakthrough that has made the solution of the present invention feasible.

[0014] (1) Development of high-precision geographic information technology In recent years, high-precision geographic information system (GIS) technology has experienced significant progress, especially in the three aspects of data acquisition, data analysis and data presentation. Rapid modeling is the basic technology that action simulation relies on. With the rapid modeling technology of accurate geographic information systems, not only can the latest geographic information data be quickly entered into the system, but it can also support the joint action simulation system to achieve dynamic environmental interaction. In terms of data acquisition, the progress of GIS technology is mainly reflected in the following areas:

[0015] First, oblique photogrammetry: This technology can quickly acquire 3D elevation information of surface objects over large areas. Compared to traditional vertical aerial photography, it provides richer elevation information, enabling 3D scene modeling and feature measurement that is close to the actual surface. Second, close-range photogrammetry: This technique captures surface information from close-up photography. It provides higher-precision geographic information data and is particularly suitable for the measurement and modeling of fine structures such as buildings and cultural relics. Regarding data analysis, advances in artificial intelligence (AI) technology enable users to conduct more complex and multi-dimensional spatial queries and analyses. This includes gaining a deeper understanding of geographic data, discovering geographic connections, and exploring geographic patterns, thereby better supporting decision-making and planning. Regarding data presentation, advancements in visualization technology, such as high-performance cloud rendering, enable GIS to provide more intuitive and dynamic map representations, helping users more easily understand and interpret geographic information. These advances are inseparable from the rapid processing of high-precision geographic data.

[0016] (2) Development of remote sensing and precise positioning technology

[0017] Remote sensing raster data is also an important component of the geographic information system that the action simulation system relies on. It is not only a true reflection of the terrain and topography, but also an important component of machine vision environment perception in this solution. The development of these technologies is briefly described as follows:

[0018] Improving Resolution: High-resolution satellite remote sensing imaging technology enables scientists to observe details on the Earth's surface more clearly. Enhancing Agility: Modern remote sensing technology emphasizes the ability to agile and maneuver imaging, which means that satellite observation angles and positions can be adjusted more quickly to respond to emergencies such as natural disasters. Improving Accuracy: With the development of high-precision positioning technologies, such as GNSS (Global Navigation Satellite System) and ground-based augmentation systems, point positioning accuracy can now be achieved from meters to centimeters and even millimeters. Achieving Intelligence and Automation: The application of AI technology in image matching, feature extraction, and expression has greatly improved the ability to obtain attribute information from massive amounts of heterogeneous remote sensing data, which is crucial for exploring the evolution of target areas. Moving towards Digitalization and Networking: The discipline of surveying, mapping, and remote sensing has achieved a transition from traditional measurement methods to digitalization and networking, which provides more efficient ways to store, process, and share data. Developing Real-Time and Popularization: With technological advances, remote sensing data can not only be obtained in real time but is also gradually being made available to the public, allowing more people to access and use this data.

[0019] (3) Development of computer computing technology

[0020] Computer computing technology has made significant progress in recent years, mainly reflected in the diversification of computing power, the acceleration of infrastructure construction, and the continuous expansion of application scenarios. Diversification of computing power: With the development of technology, computing power resources have become more abundant and diverse. This includes everything from traditional central processing units (CPUs) to graphics processing units (GPUs), tensor processing units (TPUs), and other processors designed specifically for artificial intelligence and machine learning tasks. These different types of computing resources meet different computing needs, thereby improving overall computing efficiency. Accelerated construction of infrastructure: The development of computing power has benefited from the accelerated construction of computing infrastructure, especially the construction of data centers and communication networks. Continuous expansion of application scenarios: With the rise of emerging technologies and the growth of data, the application of computing power is becoming more and more extensive, involving many industries such as intelligent manufacturing, biotechnology, and financial analysis.

[0021] (4) Explosive development of artificial intelligence technology

[0022] Artificial intelligence is developing towards general AI. This means that AI is no longer limited to specific tasks, but can handle a wider range of problems and has better adaptability and learning capabilities. The key technical points of artificial intelligence have made significant progress in algorithms, data, computing power and applications. Advances in technologies such as deep learning and machine learning have promoted the development of artificial intelligence, enabling AI to process large-scale data and complex algorithms. The high development of computing power technology has directly led to the explosive development of generative large models. In recent years, generative artificial intelligence has made great progress. Including generative large language models, generative image and video models, such as sora. All can produce the desired results through natural language descriptions. The emergence of this technology is crucial to the "natural interaction" technology in this invention.

[0023] 4. High Level Architecture (HLA)

[0024] HLA defines a simulation software architecture whose primary goal is to allow simulation applications to be implemented by combining with other simulation systems. Therefore, HLA designers must adhere to the following goals: first, they must be able to decompose a large simulation problem into multiple smaller ones; second, they must be able to combine the decomposed small simulation applications into larger ones; third, they must be able to combine small simulation applications with other unknown simulation applications to form new simulation applications; fourth, the common functionality of component-based simulation systems must be independent of specific simulation applications, and the resulting common support structure should be reusable across simulation systems; and fifth, the interface between simulation applications and common support software should be able to isolate the simulation applications from the technologies used to implement the support software, and also isolate the support software from the technologies used in the simulation applications.

[0025] like Figure 1As shown, HLA has a very complex standard specification. In August 1996, the DMSO officially published the definition, composition, and interface specifications of HLA. According to this specification, HLA consists of three main parts: rules, the object model template (OMT), and the interface specification for the run-time infrastructure (RTI). In May 2001, HLA was officially accepted as an IEEE standard, namely the IEEE STD 1516 series. 1516 addresses the HLA structure and rules, 15114.1 addresses the HLA interface specification, and 15114.2 addresses the OMT. HLA is a federated architecture. The rules section of this standard specification consists of two parts: "federation rules" and "federation member rules." HLA is an open, object-oriented architecture. Its most notable feature is that it decouples the application layer from the underlying support environment by providing general, relatively independent support service programs. This separates the implementation of specific simulation functions, simulation operation management, and underlying communication, hiding their respective implementation details. To this end, HLA provides a standard API, specifically defining detailed standards for federation member interfaces, including the services and interfaces used by federation members. These interfaces are divided into seven basic service groups: federation management, claim management, object management, ownership management, time management, data distribution management, and service support. These seven service groups describe the interfaces between federation members and RPIs, as well as the software services provided by RTI for use by HLA federation members. The standard is highly detailed and rigorous. For example, the "Federation Management" service group includes four service types. The first group includes: creating a federation execution, revoking a federation execution, joining a federation execution, and leaving a federation execution. The second group includes: registering a federation synchronization point, confirming federation synchronization point registration, announcing a synchronization point, reaching a synchronization point, and synchronizing the federation. The third group includes requesting federation save, initializing member save, starting a federation member, completing a member save, federation saved, querying federation member save status, and responding to a federation save status. The fourth group includes: requesting federation recovery, confirming a federation recovery request, starting a federation recovery, initializing member recovery, completing a member recovery, federation recovered, querying federation recovery status, and responding to a federation recovery status. The member interface specification alone includes 29 service groups, as shown in the example above. Each group contains several specific service types, and each type includes several interfaces. The HLA architecture features various structural characteristics, including hierarchical structure, data abstraction, and event-based, implicit activation. For detailed descriptions, please refer to relevant books and resources.

[0026] The general development process of HLA includes six steps:

[0027] The first step is to define the federation goal—that is, to define the ultimate goal of federation development. The second step is to develop a federation conceptual model, which abstracts the entities to be simulated and describes them using a specification language. The third step is to design the federation, which involves determining the number and functions of federation members. The fourth step is to develop the federation, which involves developing each federation member based on the federation member design. The fifth step is to integrate and test the federation, integrating the developed federation members into the federation for testing. The sixth step is to run the federation and analyze the results. These rules are defined in the Federation Development and Execute Process (FEDEP) model specification proposed by the DMSO. It provides a general, universal step for federation development, specifying the necessary processes, prerequisites, and outputs during federation development. Version 1.5 of this standard was released on December 8, 1998. This shows that the HLA simulation architecture is a fairly complete and standardized standard. Furthermore, the standard has remained relatively stable for decades, without major changes. There are already several relatively mature engineering implementations of simulation systems based on the HLA architecture, but they are mainly aimed at military and combat application scenarios.

[0028] 5. Agent-based joint action modeling and simulation methods

[0029] Let's first define the term "agent." The research and application of agent technology in the computer field originated in the 1970s, when researchers at the Massachusetts Institute of Technology (MIT) conducted a series of studies on distributed artificial intelligence (DAI). These researchers discovered that by collaborating to organize simple pieces of information into a large system, the system's ability to handle complex problems can be significantly improved. Furthermore, by defining appropriate collaborative mechanisms, the overall system's intelligence can be enhanced. In the computer field, agents are often referred to as living computer code structures that continuously and autonomously function within distributed systems, exhibiting characteristics such as autonomy, interactivity, responsiveness, and initiative. Figure 2This is a diagram of agent behavior. Agents can be categorized as thinking agents, reactive agents, and hybrid agents. The agent lifecycle consists of six states: creation, preparation, transmission, blocking, execution, and termination. In addition to supporting general operations, reducing network load, encapsulating multiple network protocols, and supporting fault-tolerant computing, mobile agents can also respond dynamically and flexibly to environmental changes. Another important characteristic of agents is autonomy: they can continue to operate without direct intervention or guidance from humans or other agents, and they can control their internal state and actions. Figure 3 This is the internal diagram of the reactive agent. Figure 4 This is a schematic diagram of the internal structure of a thinking agent. When using agent technology to model, there are three levels of conceptual abstraction: entity, object, and agent. Entity is the first level of abstraction. It is a collection of attributes, similar to a "structure" in C language. Background Figure 3The "environment" in this context can also be understood as the external world, a collection of attributes that can be specifically represented by various environmental states. An object is the product of inheritance and refinement of an entity, specifically represented as a set of attributes and corresponding action combinations. This concept is similar to the concept of an object in object-oriented programming languages, corresponding to the concept of a class in object-oriented programming languages. An object's behavior can be modeled using a mapping from the environment to actions, known as action selection. Action selection determines which set of actions should be executed next, corresponding to member functions in a class in high-level programming languages. The formal definition of an agent is an object described as having a goal, a motivation, and an internal state. A goal consists of a set of non-null attributes; a motivation is a desire or preference that leads to the pursuit of a goal, which can be understood as a "process plan and procedure description" that influences the reasoning and behavioral outputs for goal achievement. Furthermore, an agent possesses perception capabilities, action descriptions, state descriptions, and interaction descriptions. Because the concept of an agent draws on the object-oriented (OO) theoretical framework, including relationships between meta-agents (roots) and agents, such as inheritance and aggregation, I will not elaborate on these details here; please refer to the relevant literature for details. The underlying support for Multi-Agent Based Modeling and Simulation (MABMS) is the Agent Based Modeling (ABM) approach, while the engineering implementation of complex system modeling and simulation involves the design of a multi-agent system (MAS). ABM is a method that studies the interactions between large numbers of individuals or agents and the macro-level behaviors exhibited by these interactions. The ABM approach abandons the traditional way of thinking that views system modules as passive, unchanging functional modules, emphasizing the autonomy of system modules and the description of their interactive behaviors. It exhibits significant advantages in modeling flexibility, hierarchy, and intuitiveness, making it more suitable for the modeling and simulation of many complex systems. A MAS is a loose federation of multiple agents (and objects). Each agent is autonomous, with its own goals and behaviors generally unconstrained by those of other agents. Conflicts between individual agents' goals and behaviors are resolved through coordination and cooperation. A MAS can be viewed as a system with a multi-entity, loosely coupled, and open network structure, consisting of six elements: an environment E, an agent set A, an object set O, a relationship set R, an operation set P, and an operator set Γ.

[0030]

[0031] Interactions between multiple agents in a MAS are described using uniformly structured messages, distinguished by message type and content. A message is an instruction from one agent requesting another or multiple agents to perform a certain process or request information in return, and can be formally described. From a software engineering perspective, inter-agent communication is no longer suitable for traditional information system communication methods, as agents are likely to be heterogeneous and distributed. Instead, message passing based on the Agent Communication Language (ACL) is being adopted. ACL is a high-level communication language that is logically independent of the application domain. ACL only expresses communication behavior, specifying the general message type and format, while the expression of the message content is left to the specific application software developer. In terms of engineering implementation, inter-agent communication carried by ACL must rely on transport-layer network protocols such as TCP / IP, HTTP, and SMTP. A commonly used specific communication protocol, such as KQML, has approximately ten predefined types.

[0032] Current simulation system technology is mainly derived from the rapid development of network technology in the 1990s and the rise of the first generation of distributed simulation systems. As for the modeling and simulation solution technology of simulation systems, its history is even longer. Current simulation system technology has the following defects: (1) Coverage: Simulation calculations for specific aspects of specific tasks are limited by computing power factors and are not comprehensive; (2) Traditional tactical-level simulation combat units have a high degree of coupling between them, so the overall structure is strong, resulting in insufficient flexibility in unit modeling; (3) The interconnection capability of heterogeneous systems is poor, and the support capability for modern multi-service "joint combat operations" is insufficient. Summary of the Invention

[0033] To address the problems in the prior art, the present invention provides a joint action simulation method using a deterministic conditional Monte Carlo simulation scheme, comprising: a Monte Carlo simulation step: utilizing the concept of deterministic conditional Monte Carlo simulation to perform traversal calculations on key element combinations for the action target to be simulated, utilizing the high computing power of a computer system to input all reasonable combinations of key parameters into a simulation entity for calculation, ultimately automatically obtaining n results, and then screening out m optimal results; a virtual clock metronome step: designing a virtual clock metronome that coordinates various simulation modules, decoupled from real time, and used to synchronize and coordinate the simulation modules to complete tasks required in the real world within the same time interval; a perception interaction step: utilizing artificial intelligence to interact at the perception level of the simulation entity to reduce the coupling between the various simulation modules and make them reusable; an object-oriented container platform step: utilizing container platform technology to describe and encapsulate a single simulation entity or part of a simulation entity, fully utilizing and combining containerization encapsulation technology with traditional object-oriented technology to form an object-oriented container platform; and a modeling step: constructing a tactical-level action simulation system, using environmental simulation modeling as a digital foundation, and combining it with action unit simulation modeling to simulate the impact of different conditions on action results.

[0034] The beneficial effects of the present invention are: 1. Using the deterministic conditional Monte Carlo simulation scheme for tactical-level action simulation; this is a method and solver implementation invented by utilizing the advantages of modern highly developed computing power to traverse the range of a deterministic condition set and obtain N optimal solutions under different combinations of multiple conditions. The main application scenarios include tactical-level auxiliary decision-making and simulation analysis and evaluation of emergency plans, and it has broad application prospects and value; 2. Decoupling model of action simulation units based on natural interaction technology of artificial intelligence perception; the natural interaction technology proposed in the present invention is to simplify the interaction of simulation object units, reduce system coupling and improve the reusability of unit models. The inspiration and idea of ​​this technology comes from the idea that humans have only five main ways of perceiving the outside world: sight, hearing, smell, touch, and taste, but can still understand the entire world. The simulation unit itself is an AI agent, which will eventually be infinitely close to human intelligence. Therefore, a natural interactive perception solution is used to replace the software data interface commonly used in traditional computer systems. This not only simplifies the system design, but also brings it closer to the real world, making the simulation process infinitely close to the intelligent interaction in the real world. 3. Key points of the engineering design of the tactical-level action simulation solver system: a) Distributed object model and containerized encapsulation of the object model; Objects are the best technical approach to modeling things in the real world. The present invention creates a data structure of a distributed object model to describe all things, including various components and events of each simulation unit in the simulation system. The use of a distributed object model can model almost all things in the real world one-to-one, and has the ability to inherit and generalize for things of the same type. Objectification of all elements in the simulation system allows for more logical relationships between all elements (including algorithms, data, and other dynamic information) and significantly enhances module reusability. b) Virtual clock-and-metronome system and event-driven model: To simplify the complexity of interactions among simulation unit objects in the simulation system described herein (given that the types of such objects are innumerable and numerous), while ensuring consistent and rigorous logic in simulation applications, the present invention expresses all "activities" and "interactions" in the system using events. Furthermore, under the direction of a unified clock-and-metronome system, a "synchronous interaction" approach is employed to ensure logical integrity. In the event-driven model, object models are also transmitted, allowing for the expression of complex information, including methods for processing it. This event-driven model ensures simplicity and efficiency while maintaining flexibility. c) On-demand interaction model technology: The on-demand interaction model is a specialized method for interaction between simulation units and a virtual environment simulation system (digital electronic sandbox) based on a high-precision geographic information system.This new approach solves the interaction issues between simulation units and their virtual environments, simplifying and standardizing simulation system development. d) Interactive high-precision geographic information system technology: This key innovation addresses the drawback of traditional geographic information systems, which can only display static data. It enables the simulation of the interaction between simulation units and the environment within the simulation system, more realistically reflecting the dynamic nature of the real world. e) Asynchronous visualization playback: This key innovation allows for the full devotion of fewer computing resources (computing power) to the simulation process, eliminating the computational overhead of large visualization displays and the resulting slowdown. Simulation results, in the form of executed command scripts, are initially stored in a visualization buffer. This buffer is a first-in, first-out (FIFO) queue. The device responsible for visualization rendering reads the corresponding visualization commands from the output of this buffer. The visualization rendering module executes the visualization commands at its own pace, producing the visual output. Because of the buffer, even if some simulation processes take too long, the continuity of the visualization output is not affected. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is the basic HLA architecture diagram. Two members also need to communicate through RTI; Figure 2 It is a schematic diagram of the Agent's behavior; Figure 3 It is a schematic diagram of the internal structure of the reactive agent; Figure 4 This is the internal diagram of a thinking agent; Figure 5 This is a schematic diagram of an example of line-of-sight analysis; Figure 6 It is a triangulated polygonal terrain; Figure 7 This is the original schematic diagram of asynchronous rendering; Figure 8 Schematic diagram of the deterministic Monte Carlo simulation solution process of the present invention; Figure 9 This is a schematic diagram of the interaction of the virtual clock metronome of the present invention; Figure 10 is a schematic diagram of the on-demand interaction model and visual field image of the present invention; Figure 11 It is a schematic diagram of the interaction between the simulation entity and the environmental grid system of the present invention; Figure 12 This is the overall architecture diagram for the engineering implementation of the present invention; Figure 13 It is the event information flow diagram of the present invention; Figure 14 This is the MQ cluster and high availability deployment diagram of the present invention; Figure 15 It is a system deployment diagram of the present invention. DETAILED DESCRIPTION

[0036] The goal of the present invention is to use the computing power of a computer system to simulate and analyze actions that are about to be carried out and that depend on time and geographical environment, so as to find the best action plan (the optimal solution to the problem), thereby making the action have a higher success rate.

[0037] The main problem solved by the present invention is: a simulation scheme and its dedicated solver invented in order to "enable actions that rely on time and geographical environment to achieve a relatively high success rate under the influence of constantly changing external factors." The external environment mentioned above includes but is not limited to: equipment performance, personnel capabilities, weather influences, geographical environment, and the impact of their changes over time. By performing deterministic conditional Monte Carlo simulation on these external environmental factors, relying on powerful computer computing power to calculate (traverse) the simulation results of actions under various situations, the best solution is found. This is actually a complete Markov decision process of action, which refers to a decision-making method based on the Markov Decision Process (MDP), which is a mathematical (engineering) framework that indicates that machine learning systems can autonomously explore the environment and learn how to expect to obtain the maximum reward in the future.

[0038] The innovative focus of the present invention lies in how to conduct "joint" action simulation for multiple groups belonging to different organizations and relying on different technologies, and obtain results with higher confidence.

[0039] For example, during disaster relief efforts, transporting supplies requires the fleet to reach the disaster site as quickly as possible. Before embarking on an operation, careful consideration must be given to factors such as the route, the impact of rain on the road, visibility, wind speed, the load-bearing capacity of bridges, fuel consumption, crew fatigue, and the impact of the disaster on the geographical environment. Furthermore, weather changes, such as rain at different times during the operation, can have varying impacts on the operation. The steepness of the terrain can affect fuel consumption, and the distribution of gas stations can be crucial factors in transport operations. Some of this information is static, some is dynamic, and some is interactive. For example, an uphill dirt road in a dry country area may be passable for vehicles of a certain load capacity. However, in light rain, specific tires may be required to make the road passable. Heavy rain may make the road completely impassable or restrict the load capacity—these details can be crucial to the success or failure of an operation.

[0040] All parameters are input to the solver, which performs calculations along a timeline, yielding different results. By selecting reasonable scenarios, there are tens or even hundreds of thousands of possible combinations of parameters. For each possibility, the solver performs calculations, drawing conclusions about key factors influencing the operation. These conclusions can serve as a basis for commanders' operational decisions and can also be used as a backup for a tiered contingency plan library.

[0041] This type of tactical-level action simulation depends not only on the level of detail in unit modeling but also on high requirements for extensive data organization and hybrid computing capabilities. This invention falls under the category of "Joint Operation Analytic System," primarily used for pre-action simulation analysis, with a focus on maneuver modeling, environmental modeling, and perception modeling. It's easy to see that this "pre-action simulation" calculation is essentially an "analysis" of the feasibility of the planned action. It can effectively help managers formulate tactical-level action plans and analyze the impact of various possibilities on the action in advance. This can significantly reduce the serious consequences or even casualties that can arise from inconsiderate considerations during actual actions. This can significantly reduce operational risks, save associated operational costs, and allow commanders to be more informed, thereby achieving a higher success rate. Of course, the higher the confidence level of the simulation results, the greater the number of parameters required, the greater the computing power consumed, and the longer the time. It should also be noted that simulation results serve as auxiliary decision-making materials for action commanders and cannot replace their decisions.

[0042] This invention is primarily used in scenarios such as action planning analysis that are highly dependent on geographical conditions and time. The term "action" here refers to scenarios such as marches and battles, disaster relief, and long-distance transportation. Before a specific action is launched, parameters influencing potential external factors can be input to generate different outputs. If the results are unsatisfactory, controllable "planning parameters" (such as departure time, equipment used, and personnel arrangements) can be adjusted and the simulation repeated to achieve better results.

[0043] "Joint" is another key word. It not only expresses the parties involved in the action, but also refers to the parties involved in the simulation analysis. The latter can be understood as the "subsystems" involved in the analysis, such as: vehicle simulation subsystem, weapon simulation subsystem, or weather simulation subsystem. It can be seen that each subsystem is a different "professional field" with very different parameter algorithms. How to solve the problem of interconnection and mutual cooperation is a key issue to be solved by this invention.

[0044] In addition to providing the corresponding theoretical methods, this invention also presents a system engineering architecture design for a dedicated solver for this solution. With certain adjustments, this solution and its solver system can be widely applied to a wider range of dynamic planning application scenarios based on precise geographic information systems (digital electronic sandboxes), such as engineering construction, urban planning, and urban emergency response. The core technology used in this invention is highly dependent on computing power. Only with today's highly developed computing power can we break away from traditional information architecture models and implement new solutions. In other words, many traditional simulation systems were born in an era of scarce computing power, forcing them to adopt compromised design solutions. This invention fully leverages artificial intelligence (AI) technology to simplify system-to-system interactions, including generative AI achievements such as virtual machine vision, sound recognition, and large language models. This fully decouples the various modules in the action simulation system, using a "natural interaction" approach that more closely resembles the interactions between real systems (entities). This is a new solution that was only possible after significant advances in AI technology. While the key technical core and main architecture of this invention share similarities with the MAS solution, they exhibit significant differences in mathematical principles and engineering implementation. Specifically, its innovations are reflected in the interaction mechanism between entity agents, the microstructure of the entity model, the collaboration strategy between entities, the organization mode of data, and the modeling process.

[0045] The present invention addresses the defect (1) of the current simulation system technology in the background art by introducing a deterministic conditional Monte Carlo simulation scheme; addresses the defect (2) by introducing a natural interaction mechanism; addresses the defect (3) by introducing an AI Agent simulation algorithm and a universal solver information architecture implementation scheme, which has strong security and confidentiality capabilities (for parameters and algorithms).

[0046] Overview Summary: This invention is a universal solution for traditional tactical-level action simulation. It has the advantages of high flexibility, large coverage area (bringing higher confidence in the results), and strong versatility. It can standardize existing action simulations and fill the gap in related fields in China.

[0047] The following will compare and contrast the differences in system design engineering methods between the present invention and existing simulation technologies. It should be noted that this discussion does not include traditional stand-alone simulation technologies (such as wargaming), but rather focuses on the situation after the emergence of distributed simulation systems.

[0048] 1.1 Modeling technology of existing simulation systems and its optimization and improvement

[0049] 1.1.1 Shortcomings of the Existing Technology The models in traditional simulation systems include physical models, mathematical models, program models, random models and logical models. Here we only discuss mathematical models and program models. Mathematical models are represented by algorithmic processes and mathematical equations. These models consist of a series of mathematical equations or relationships, and their discrete solutions can be obtained. Usually, the model uses numerical approximation methods to solve complex mathematical functions that have no specific solutions. The program model is a form of expression that describes the dynamic relationship of situations, which are expressed in mathematics and process logic. These models are usually called simulations. Therefore, in the traditional simulation model system, it is generally believed that each mathematical model is represented by a certain function, and the curve of this function is required to be as close as possible to the actual observation results, which can be used as an indicator.

[0050] As mentioned above, in traditional simulation systems, the closer a mathematical model is to the real world, the better the design. However, such models typically produce deterministic results given the same input parameters. Therefore, to further approximate real-world conditions, traditional simulation system modeling schemes often require intentional interference through the addition of "random factors." However, determining which parameters to interfere with and how much interference to apply is a complex issue. This is closely related to the complexity of the mathematical model (for example, a function with thousands of parameters) and the simulation object. Each model is unique and requires individual attention.

[0051] 1.1.2 Objectives and Optimizations of the Invention

[0052] The present invention provides two innovative ideas in the simulation modeling scheme.

[0053] (1) Use the Monte Carlo simulation method to perform macro-deterministic conditional traversal. It does not focus on the functions and parameters of the mathematical model of the specific simulation object itself, and encapsulates the mathematical model to form an object (Object). By traversing the macro-indicators (finite set) of the external environment, according to the overall idea of ​​finite element numerical analysis, the parameter input and simulation calculation of the model object are realized. (2) Use the artificial intelligence model trained by neural network deep learning to replace the traditional "deterministic" mathematical model. This idea is to make full use of the "incomprehensibility" principle of the neural network training model to replace the original "deterministic parameters + random interference" traditional model. The main purpose of this replacement is to greatly simplify the complexity of model parameters and interactions.

[0054] Using the idea of ​​the present invention, for example Figure 5 As shown:

[0055] In traditional perception modeling, line-of-sight analysis is a common simulation computing requirement. It primarily addresses the question of whether points A and B can see each other in a 3D environment amidst uneven terrain (because light travels in straight lines).

[0056] Because three-dimensional geographic information systems (3D-GIS) typically use vector data to represent terrain, common "line-of-sight analysis" algorithms interpolate between terrain elevation points using a three-dimensional vector model of the terrain. If the line connecting the observer and the target is above all key elevation points, then there is line of sight; otherwise, there is no line of sight. This is a standard analytical set mathematical calculation model. However, such a scheme must consider many other factors, such as the observer's height, the presence of vegetation, the density of vegetation in different seasons, the curvature of the Earth's surface, and the size and color of the observed object (for example, a red tank is obviously easier to spot than one painted in camouflage). Furthermore, dynamic weather conditions such as fog, dust, and rain can also affect line of sight (with varying degrees of rain affecting the target). As can be seen, the number of these "external factors" is simply impossible to fully enumerate (infinite). While computationally expensive, their complexity and lack of generalizability are also significant factors that affect simulation results. This is because not everything in nature can be accurately described by mathematical models. This is due to the "open" nature of complex systems. Therefore, to address this issue, the present invention summarizes the aforementioned problems into a single macroscopic problem: "vision." A model of the observer (e.g., object A) is assigned to a machine vision algorithm (perhaps trained to simulate the human eye or an observation instrument). Using A's coordinates, a perspective image or video of the three-dimensional scene within its field of view is obtained, just as if observed by the human eye. In this way, no matter how complex the external environment, the "environmental modeling" subsystem is solely responsible. For example, to simulate the impact of rain on observation behavior: at a macroscopic level, rainfall levels might be categorized as 1-10. Simply superimpose the effects of different rainfall levels on the image within the observer's field of view. This fully decouples the complexity of the observer from that of the environment, transforming countless environmental variables into a single visual effect. To calculate "which rainfall level results in the worst-case scenario," simply iterate through the rainfall levels from 1 to 10 within the environmental modeling system and repeat the calculation. This effectively transforms "uncountable factors" into the "finite set of parameters" mentioned above. It's worth noting that the machine vision model proposed in this paper, based on neural network training, is scenario-independent and reusable after training, significantly reducing overall system complexity. Furthermore, due to the "incomprehensibility" of large neural network model algorithms, "hallucinations" or "cognitive biases" (commonly known as "recognition errors") may sometimes occur for the same visual scene. However, such "errors" perfectly simulate the randomness of human observers.

[0057] From the above examples (examples of machine vision perception), it can be seen that this method can be easily extended to other "alternative interaction schemes using large neural network models". Just like the recognition of natural pictures by machine vision models; the recognition of natural sounds by auditory models; the recognition of natural simulated radio wave field strength by radio reception models, and so on. Therefore, the most critical optimization innovation of the present invention is called the "natural interaction technology" scheme. This scheme simplifies the interaction factors that cannot be "exhaustively enumerated" in traditional simulation systems into more macroscopic interaction factors. Use a large neural network model trained by deep learning to replace the traditional function row mathematical model. For example: the "observer"'s external perception model is simplified to "vision" and "hearing" according to the application scenario; the model of the "engine" operating condition affected by external factors is simplified to the external environment "temperature" and "air pressure", etc.

[0058] Traditional entity simulation granularity often results in "infinite layers of recursion." The concept of "simulation boundaries" is often used to address granularity issues, forcibly truncating recursion. The present invention's macroscopic natural interaction modeling method effectively transforms this into a "finite set," further decoupling the simulation entity model from the external environment and providing sufficient universal reusability, significantly improving the fidelity and efficiency of the simulation system while significantly reducing complexity.

[0059] This innovative point of the present invention has been fully verified in the process of technological development of natural language processing. The technical solution of natural language translation has been transformed from the earlier attempt to base on grammatical analysis and traversal to the current LLM (large language model) trained with massive corpus. The former can be understood as an attempt to mathematically model natural language, but it was soon discovered that the complexity of language could not be exhaustively enumerated; while the latter uses a process similar to that of human language learning to "naturally understand" the semantics. It turns out that the latter uses a universal method to replace the previous infinitely complex solution process. Despite the "incomprehensibility" of its solution process, it has been proven to be the most effective solution.

[0060] Similarly, this approach is used in the present invention to address the complexity of the mathematical modeling process for simulation systems. The tactical (action) level simulation system modeling types targeted by this invention primarily include: environmental modeling, mobility modeling, perception modeling, effects modeling, and communications, command, and control (3C) modeling. In addition to the perception modeling described in the preceding example, other modeling approaches can also be implemented using the above approach.

[0061] 1.2 Scenario Algorithms of Existing Simulation Systems and Their Optimization and Improvement: Scenario algorithms are crucial to simulation systems. In a simulation system, a scenario is a setting or assumption about factors such as events, environment, and entity behavior during the simulation process. Simply put, it is a detailed "script" for how the simulation will proceed. For example, in a military simulation system, the scenario will include the geographical environment of the battlefield (such as mountains or plains), the number and performance of the participating troops and equipment, the combat mission (whether it is offense, defense, or reconnaissance), possible events (such as encountering enemy reinforcements or sudden weather changes), and many other aspects.

[0062] 1.2.1 Types and Disadvantages of Existing Simulation Scenario Technologies

[0063] (1) Case-based scenario setting: construct scenarios with reference to actual events. For example, in the simulation of aviation accident emergency drills, similar flight conditions (such as weather conditions, aircraft models, fault types), airport environments (runway conditions, rescue facility locations) and personnel responses (crew response measures, ground rescue personnel actions) are set based on previous aviation accident cases. The advantage of this approach is that it is close to reality and can effectively simulate actual problems; the disadvantage is that it may be limited by the specific circumstances of existing cases and lacks exploration of new situations. (2) Parameterized scenario setting: create scenarios by setting the range and change rules of various parameters. Taking power system simulation as an example, parameters such as grid voltage and load power can be set to change according to a certain functional relationship within a certain range, and parameters such as the output power response strategy of the generator can be set at the same time. This allows for flexible study of the behavior of the system under different parameter combinations. However, this method is highly coupled and requires a deep understanding of the relationship between system parameters, otherwise parameter combinations that do not conform to reality may be set. (3) Rule-based scenario setting: construct scenarios according to pre-established rules. For example, in urban traffic simulation, vehicle driving rules (such as obeying traffic lights and keeping a safe distance) and pedestrian crossing rules are formulated, and then different traffic flows, road facility changes, etc. are set according to these rules. This solution is convenient for systematic simulation of complex systems, but the formulation of rules may be relatively complex and need to be constantly updated to adapt to changes in actual conditions. (4) Hybrid assumption setting: a combination of the above methods. For example, in military simulation, the basic combat scenario (geographical environment, participating forces) is first determined based on actual war cases, and then the range of variation of weapon performance parameters is set using parameterization methods, and then the combat strategies of both sides are set according to combat rules. This method combines the advantages of various schemes and can construct simulation assumptions more comprehensively and flexibly, but it has high requirements for designers and requires comprehensive consideration of the coordination of multiple factors. Moreover, the designed simulation system is too "specialized" because it is not reusable (that is, it is almost impossible to apply it to another simulation task in a similar application field, and even modification is difficult). Therefore, the cost of such a simulation system developed specifically for a certain task is too high. It can be seen that the simulation goal for complex tactical-level tasks is to "get as close to the real scene as possible." However, this often faces trade-offs between many factors such as effect, cost, performance and complexity, making it difficult to be comprehensive.

[0064] 1.2.2 Objectives and optimization improvements of the present invention: There is a concept of "simulation depth" in simulation systems. Simply put, simulation depth refers to the degree to which the simulation restores the details and behaviors of the real system. If the real system is regarded as an object with multi-level, complex internal structure and behavior, the simulation depth is like the degree to which the internal details of this object are excavated. For example: in an aircraft flight simulation system, a shallower simulation depth may only simulate the basic flight trajectory of the aircraft, such as the approximate path of takeoff, cruising and landing, and the simulation of the internal system of the aircraft only involves simple power output and basic control signals. In deep simulation, each system of the aircraft will be simulated in detail, including the internal combustion process of the engine, the delicate operation of the hydraulic system, the complex signal interaction of the avionics system, and even the specific behavior of the aircraft under various fault conditions, such as how the aircraft's autopilot system responds when a sensor fails.

[0065] In the field of military simulation, shallow simulations may simply present the general course of a battle, such as a simple confrontation between the forces of both sides and the outcome (for example, the ancient Lanchester model based on partial differential equations). Deep simulations involve detailed details of combat unit operations, such as individual soldier tactical maneuvers, the precise strike effects of weapons and equipment, and the specific impacts of complex terrain and weather on combat. Obviously, the deeper the simulation depth, the greater the cost and other "overheads." The present invention aims to resolve this contradiction, specifically in the design of simulation "scenario elements." It attempts to reduce system design and operation costs by improving the reusability of "scenario element modules." In other words, by enabling "scenario element" modules (or "subsystems") to be reused for multiple different types of missions without requiring redevelopment or extensive modification, the overall system simulation depth is maximized without increasing costs. Achieving this goal relies on a core guiding principle throughout the present invention: "decoupling." That is, the coupling between the various modules of the scenario element is minimized. For example:

[0066] For example, in environmental simulation, factors such as wind, rain, cloud, fog, temperature, and light are all independent subsystems within the same grid coordinate system. These weather elements are solely responsible for calculating the "performance value" for each grid. The integrated weather grid system is responsible for uniformly calculating the "combined effect" formed by these multiple values, such as visibility (combining rain, fog, and light). The visualization system then displays this combined effect. Similarly, in perceptual simulation, the propagation of sound and light are independently calculated and ultimately "displayed" as "effects." Other simulated entities that require environmental factors only need to respond to these "effects." This design significantly reduces the coupling between modules. With this design approach, each module only needs to perform calculations based on its own simple or complex mathematical model and then assign an "effect value" to each coordinate grid, without considering the influence of other modules. For example, consider a scenario where sound intensity at a source is 100% and diffuses outward through the air. For a point sound source, the sound waves it emits propagate evenly in all directions, gradually diffusing the energy. The sound wave energy per unit area decreases with increasing distance from the source and is inversely proportional to the square of the distance. In other words, for every doubling of the distance from the sound source, the sound intensity (sound strength) decreases by 1 / 4. Converted into decibels, this attenuation is approximately 6dB. This is a simple mathematical model. However, when simulating a real environment, there may be many different sound sources. These sources all calculate the sound attenuation of a specific grid according to this model, and the grid system produces a superimposed sound effect. The simulated entity only needs to "listen" to this combined effect, which is the closest to the real situation. Similarly, this approach also applies to radio waves and other objects.

[0067] The present invention also takes into account more complex situations. For example, when there are obstructions, the degree of sound attenuation will also be different. This is still calculated by the simulation system according to the specific situation of each grid, and finally presented through the physical effect. However, the complexity of the sound environment module will not increase any more - it only needs to calculate the attenuation of a certain frequency of sound in each grid, and the factors that need to be considered in this calculation are "enumerable". And such an independent hypothetical module is universal, and its hypothetical parameters are also limited. This is also a manifestation of the use of "first principles" in the present invention. The biggest difference between this and the existing hypothetical system design ideas is that external interactions use "comprehensive effects" instead of using multiple parameters to interact.

[0068] 1.3 Existing mainstream distributed simulation system technology architecture and its optimization and improvement

[0069] This section primarily describes the optimizations and improvements in system architecture design presented in this invention, applying a new information technology architecture to simulation system design. 1.3.1 Shortcomings of Existing Technology: Existing simulation frameworks are primarily based on distributed computing frameworks. The following briefly analyzes several popular simulation system architectures.

[0070] (1) High-Level Architecture (HLA): The High-Level Architecture (HLA) simulation system architecture is a distributed simulation framework developed in the 1990s. Its design goal at the time was to enable joint simulations between different organizations. However, with the rapid development of information technology, it has gradually exposed its backwardness, as shown below:

[0071] First, development is complex. The HLA system itself is complex, and developing and maintaining simulation systems requires developers with high technical skills and extensive experience, resulting in a high learning curve. All simulation systems participating in a simulation federation must integrate with the HLA Runtime Infrastructure (RTI) based on a set of interface standards. This is the core of the HLA simulation system, providing a series of standard interface services for simulation applications. RTI is an interaction method based on data protocols, so variations in protocol interpretation are a significant issue. RTI primarily provides six management services: federation management, declaration management, object management, ownership management, time management, and data distribution management. These services enable interoperability and collaboration between different simulation models (federation members), such as controlling the entry and exit of federation members (federation management), managing the declaration and discovery of object instances (declaration management), and handling object instance updates and reflection (object management). Second, operational efficiency is low. Because its architectural design prioritizes universality and interoperability, operational speed may be affected in scenarios with extremely high real-time requirements. For example, in flight simulation training scenarios with high real-time requirements, latency may occur. Third, compatibility with standard updates is essential. As technology evolves and HLA standards are updated, existing simulation systems may face compatibility issues, requiring resource-intensive updates and adaptations. Due to the inherent complexity of HLA, developing interoperability interfaces is complex, leading to inconsistent interpretations of the standard.

[0072] (2) Model-Driven Architecture (MDA)

[0073] The Model-Driven Architecture (MDA) is a simulation system architecture based entirely on object-oriented thinking. It boasts advantages such as a high level of abstraction, platform independence, high development efficiency, and good portability. By establishing a Platform-Independent Model (PIM), MDA abstracts simulation logic from underlying technical details. By emphasizing model separation, the transition from PIM to PSM makes it easier to port software systems between different software platforms or hardware environments. The software development process revolves around models, driven by their refinement and transformation. These models can better facilitate communication between different roles, such as developers and simulation analysts.

[0074] However, MDA also presents the following challenges: First, model conversion is complex: The conversion process from PIM to PSM is quite challenging, especially for complex simulation systems. Complex conversion rules and tools need to be defined, and these rules are difficult to apply universally, requiring customization for different domains and platforms. Second, tool support is limited: While some MDA tools exist, they are not yet mature. These tools may experience performance issues or lose key model attributes when handling large and complex models, or provide incomplete modeling support for specific domains. Third, the requirements for personnel are high: developers need to master both modeling knowledge and target platform knowledge. For example, developers must be proficient in modeling techniques such as the Unified Modeling Language (UML) and understand the characteristics of the hardware platform and operating system on which the software will run.

[0075] (3) Web-based architecture (XMSF)

[0076] In the early 2000s, the development of Web services technology represented the next phase in distributed computing. The Extensible Modeling and Simulation Framework (XMSF) was defined as a set of standards, descriptions, and recommendations for Web-based modeling and simulation. XMSF leveraged XML markup, Internet technology, and Web services to facilitate the application of this simulation architecture. Clearly, this was the next generation of distributed computing architecture applied to the traditional simulation field following the emergence of Web services. This architecture introduced XML primarily to address the issue of updating proprietary protocols, such as the complexity of the RTI interface specification in the aforementioned HLA architecture. Because XML is extensible, it provides a self-explanatory foundation for flexible interactions between simulation entities. Furthermore, the standardization of Web services, with its relatively low development threshold, addressed the development difficulties of the two aforementioned simulation architectures. However, while inheriting the advantages of the Web services architecture, it also inherited its disadvantages. First, Web services inherently lack real-time responsiveness, and their text-based protocols are relatively inefficient, making them less suitable for real-time simulation. Second, XML's extensible markup language has a high parsing overhead, and XML itself has relatively vague data type definitions. For example, the value of an element can be a number, a string, or any other type, but it does not distinguish strictly like programming languages. This may lead to type mismatches or misunderstandings of data during data processing and conversion.

[0077] (4) Multi-agent based simulation modeling (MAS)

[0078] The architecture of a multi-agent (Agent) modeling and simulation system has been described in detail previously and will not be repeated here. The characteristics of MAS are briefly summarized as follows: (a) Autonomy: This includes self-management and self-regulation. (b) Distribution: This includes structural distribution and computational distribution. (c) Interactivity: This includes interaction between agents and with the environment. (d) Collaboration: This includes task decomposition and allocation, as well as joint decision-making. (e) Adaptability: This includes environmental adaptation and learning capabilities. (f) Complexity: This includes complex system behavior and complex model construction.

[0079] The last point also mainly reflects the shortcomings of this architecture in the current environment.

[0080] Complex system behavior: Due to the interactions and collaboration between multiple agents, system behavior often exhibits complexity and uncertainty. The overall system behavior is not simply the sum of individual agent behaviors, but rather emergent behavior resulting from the interactions and influences between agents. Complex model construction: Building a simulation system based on multi-agent technology requires detailed design and modeling of agent behaviors, interaction rules, and environmental models, which increases system complexity and development difficulty. Developers require solid knowledge and skills in artificial intelligence, distributed computing, and system modeling to effectively build and apply multi-agent simulation systems.

[0081] Here, the shortcomings of the MAS architecture are briefly described in relation to the specific content of the present invention.

[0082] First, most current multi-agent technologies are still based on traditional knowledge bases and expert system technologies. This type of simulated entity processes external events in a deterministic and understandable manner. This makes it difficult to approximate real-world simulated entities at the individual level. Second, current multi-agent simulation modeling architectures lack large-scale engineering implementation technologies and are difficult to achieve universal application. Third, there is the issue of communication complexity. In a multi-agent system, agents need to communicate frequently to exchange information, coordinate actions, and reach consensus, and proprietary protocols and middleware are required for interaction. However, as the number of agents increases, the complexity of communication increases exponentially, potentially leading to communication delays, information congestion, and communication failures.

[0083] It can be seen that the simulation system is generally more than 10 years behind existing technologies in terms of information system architecture implementation technology. For example, the HLA architecture is a remote inter-process interaction technology from the 1990s. Especially when it comes to heterogeneous interaction, RTI also uses complex and outdated technologies such as CORBA. In addition, its important part, the RTI interface, adopts a standard C++ / Java interface: RTI provides standard C++ and Java language interfaces. This makes the threshold for developers to develop simulation systems very high. Therefore, applying new software system implementation technology to the realization of simulation systems is an important goal of the present invention.

[0084] 1.3.2 Objectives and optimization improvements of the present invention: Information technology is changing with each passing day, and new ideas are constantly emerging in the system technology architecture. In view of the shortcomings of the aforementioned traditional simulation system implementation architecture, the present invention makes the following improvements in the computing architecture. (1) Use a microservice container object architecture to carry the new generation of simulation systems. This new technical architecture greatly reduces the coupling degree of each simulation object. They can be heterogeneous systems developed by different developers, using different tools, and different development languages ​​and frameworks. Container technology isolates the content from the external environment and has a unified interface method. This can make full use of existing achievements, especially professional algorithms developed by non-information technology personnel. For example, the algorithm for the propagation of radio waves in the atmosphere developed by radar experts (who are not professional programmers) is likely to be implemented using their dedicated simulation tools, such as Matlab (a mathematical calculation and graphics tool software). Under the container platform, only encapsulation is required, rather than rewriting. This greatly improves the ability of heterogeneous systems to serve each other. The essence of the simulation system is to study the relationship and behavior of different independent simulation entities.

[0085] This independent container object model architecture corresponds well to a large number of simulation entities. Specifically, each simulation entity can be easily and naturally mapped to one or more container object models. Furthermore, the entities hosted by these container object models can be heterogeneous. Traditional architectures, however, are not easy to achieve such fine-grained interaction between heterogeneous objects. This natural correspondence, at its underlying logic, is well-suited to the overall architecture of such systems. A tactical simulation system built on this logic will manifest as a structure composed of thousands of interacting container objects, each representing a simulation entity or a portion of one. Furthermore, this is completely decentralized.

[0086] (2) A sequence message mechanism is used to complete the communication between simulation modules and simulation entities. The aforementioned current traditional simulation systems (here specifically referring to "distributed simulation systems") mostly use prescribed proprietary protocols for the interaction between simulation objects. For example, the communication between HLA's RTI and each simulation module uses the traditional API interface to transfer data; and XMSF is based on the HTTP protocol for interactive communication. The problem with the former is that it is complex, difficult to develop, and lacks flexibility; the problem with the latter is that reliability cannot be guaranteed. The present invention uses a universal and reliable message sequence (MQ) protocol to carry content between simulation entities. The advantages of its improvement are as follows:

[0087] Decentralization. A central business node, like the RTI dispatching system, is not required to manage all business data. The message transmission system is independent of the business. All simulation entities (such as agents) that need to communicate can communicate freely point-to-point (PP) within a common channel. Flexibility. Different types of simulation entities can interact through the MQ messaging system using their own protocols, without having to adhere to a set of common business-related rules. This facilitates the full utilization of existing achievements by different organizations to jointly build a large-scale simulation system. If both endpoints needing communication belong to the same organization or subsystem, they do not need to modify their communication protocols. Furthermore, modifications to their own "small circle" protocols will not affect the overall system. Timing reliability. The most important feature of a simulation system is the timing reliability of "events." It is necessary to ensure that the "target is hit first" and then the "projectile flies over." The order of events is often a key issue that requires significant effort to resolve in simulation systems. In the continuous reliability message queue used in this invention, each message corresponds to a specific "event" from one entity to another point-to-point. Each message has its own timestamp, and the message queue mechanism ensures the order of the messages. This prevents the "last sent first" or "first sent last" message from arriving due to network instability or other reasons.

[0088] From the above, it can be seen that the most important improvement of the present invention is that the use of a new information architecture makes it easier to build a complex simulation system.

[0089] (3) A high-precision three-dimensional geographic information grid system is used as a unified data base to support the operation of the matrix system.

[0090] The simulation system designed by the present invention is mainly aimed at the application scenario of optimizing the action planning at the tactical level. Including but not limited to combat operations, disaster relief and other types of action tasks. This type of simulation task has characteristics that are closely related to topography, climate environment, etc. Therefore, the three-dimensional grid system carried by the high-precision geographic information system is used in the present invention as the data base of the entire simulation system. The present invention uses a three-dimensional grid system, which can be understood as a spatial system composed of small cubes (or called spatial lattices, which can be regarded as pixels in a three-dimensional game scene). Its resolution can be adjusted, and the specific value depends on the resolution required by the simulation task.

[0091] The 3D GIS itself is a large data structure composed of vector data. For example, the traditional 3D terrain surface is composed of many triangular facets, such as Figure 6As shown, this three-dimensional grid system serves as a unified data base for all simulation tasks, providing a unified spatial standard for the numerous simulated entities within the task. The system implementation designed by the present invention differs from the principles of existing traditional simulation systems: the latter typically treat environmental simulation as an independent subsystem on a par with other simulation systems, while the former treats the geographic environment system as a platform for carrying and exchanging common data. Each point in its grid is a complex data structure object, superimposing the "combined effects" of the various subsystems acting on that point. For example, a geographical point may have a "friction coefficient," but this coefficient is calculated based on factors such as "time + location + surface hardness + rainfall amount + temperature." When a vehicle passes through this location, this "friction coefficient" is factored into the calculation of the vehicle's passability. As can be seen from the above example, the data base itself is dynamic, participates in the calculation, and, most importantly, serves as a transit point for data exchange. All simulated entities are aligned within this unified spatial standard, with a unified time cadence, to accurately simulate and map the real-world situation.

[0092] (4) Use “asynchronous visualization” technology to present the simulation process.

[0093] The simulation results are ultimately presented. Monte Carlo simulations, in particular, generate numerous simulation results, allowing simulation personnel to evaluate and select the optimal results to confirm reasonable action plans and solutions. For the application scenarios to which the present invention is applicable, the visualization of simulation results does not require the simultaneous real-time performance of the simulation process. In other words, the simulation process is essentially a fully automated calculation process with no human intervention, and the visualization of the final results merely serves to "play back" the simulation process.

[0094] The present invention adopts a solution called "asynchronous visualization" in the presentation of simulation results. The biggest advantage of this solution is that it can put fewer computing resources (computing power resources) into the calculation of the simulation process, and will not slow down the simulation process because of the large amount of visualization that occupies computing resources. Visualization requires a lot of image rendering calculations. However, in the simulation scenario using the Monte Carlo simulation solution, it is not necessary for every simulation process to be visualized in real time. What ultimately needs to be "replayed" are often several preferred solutions selected by the simulation staff. And most of the simulation results will eventually be discarded.

[0095] This visualization of the simulation process effectively utilizes computing resources. It allows relatively complex simulations to be run on hardware platforms with limited computing power. Because this process is a fully automated Monte Carlo simulation, no human intervention is required. Therefore, simulations can be performed during periods of relatively abundant computing power, such as nighttime and holidays, to fully utilize computing resources. After the simulation is complete, the operator can invoke the visualization module to replay and analyze the results of interest. Figure 7 This is the original schematic diagram of asynchronous rendering.

[0096] Another meaning of "asynchronous" playback is that the effects produced during the simulation, in the form of executed command scripts, are first stored in a visualization buffer. This buffer is a first-in, first-out (FIFO) queue. The device responsible for visualization rendering reads the corresponding visualization instructions from the output of this buffer. The visualization rendering module executes the visualization instructions at its own pace, producing the visual output. Because of the buffer, even if some simulation processes take too long, the continuity of the visualization output will not be affected.

[0097] 2.1 Prerequisites for the Technical Solution

[0098] The application scenarios and constraints to which the present invention is applicable are stated here.

[0099] The present invention is not a real-time simulation system (such as flight simulator software), but rather a solver for optimal action planning. Therefore, the virtual clock metronome of the simulation system described in this invention is not required to be consistent with a real-world wall clock. The specific concept is as follows: In the simulation system described in this invention, internal virtual time and external real time do not necessarily correspond; they can be faster or slower than real time. During the simulation of an action, the time scale is slowed down or sped up to facilitate the use of a wider range of conditional parameter combinations in the calculation. For example, a 10-hour action may have 1,000 determinants (and their combinations) that affect the success or failure (probability) of the action. System simulation personnel will select reasonable combinations (e.g., 100,000 reasonable combinations) and try them one by one using Monte Carlo simulation methods. Ultimately, the 10 optimal results are obtained and used for manual performance evaluation. This calculation process may take 500 hours. Although time-consuming, it is a planning evaluation step before the actual action begins. Therefore, compared with the consequences of the success or failure of actual actions, the time factor of the ex ante simulation is not a key cost factor, and the system does not require any real-time performance. In the end, it is only necessary to select the optimal parameter combination simulated by the system and give it as a reference for project decision-making, which can greatly improve the success rate of the project. This is the usage method and applicable scenario of the simulation system based on the Monte Carlo method described in the present invention. It is more suitable for application scenarios such as prior rationality evaluation of emergency response plans. In general, the application scenarios faced by the present invention are simulation application targets with the following restrictive characteristics. That is, it is applicable to the evaluation process of the "Monte Carlo simulation" method. The general process is as follows:

[0100] (1) The process of a single simulation does not require "human" participation. The entire process is completed by the computer system itself, and then the simulation results are given. The process can be fast or slow, and does not need to be real-time; (2) The simulation user adjusts the parameters according to the results, or the system automatically adjusts the parameters according to the established rules, and then repeats the simulation process to obtain different results. This process is then repeated n times; (3) The n (possibly large) results are sorted and analyzed, and the optimal parameters are deduced, that is, the optimal solution for "action planning"; other simulation scenarios with real-time requirements or non-pre-evaluation types are not suitable for this invention.

[0101] 2.2 Basic principles of the key technologies of this invention

[0102] This section focuses on describing the key innovative ideas and basic principles of the technology and their implementation solutions in this invention. The content described in this section also fully encompasses the innovative points that need to be protected in this invention.

[0103] 2.2.1 Principle of Deterministic Conditional Monte Carlo Simulation Solver

[0104] Monte Carlo simulation is a computational method that uses random sampling to solve problems in fields such as mathematics and physics. Its basic idea is to use random numbers to simulate various possible scenarios. For example, in finance, Monte Carlo simulation is often used for risk assessment and asset pricing. It can simulate various possible future asset price trends, thereby helping investors assess the risk of their portfolios. The steps generally include determining the distribution of variables, generating random samples, constructing a simulation model, repeating the simulation calculations multiple times, and analyzing the simulation results.

[0105] The present invention adopts the idea of ​​"deterministic Monte Carlo simulation" or "conditional Monte Carlo simulation" (hereinafter referred to as "deterministic conditional Monte Carlo simulation"). In conventional Monte Carlo simulation, a large number of random sampling calculations are performed on random variables. In the given condition set traversal calculation scenario described in the present invention, the core method still uses the basic idea of ​​Monte Carlo simulation (approximating the result through a large number of repeated calculations), but the sampling process is not completely random, but is calculated in sequence according to the given condition set (simulation assumptions), so as to obtain statistical results under these specific conditions. This method is helpful in analyzing the behavior or characteristics of the system under certain certain conditions. The present invention uses the idea of ​​deterministic conditional Monte Carlo simulation to perform traversal calculations on the combination of key elements (deterministic condition sets) for the action target to be simulated, and uses the high computing power of the computer system to input the reasonable combination of all key parameters within a reasonable range into the simulation entity for calculation, and finally automatically obtains n results, and then screens out m optimal results from them. This calculation process is completely automatic and does not require human intervention. It is mainly used in application scenarios for pre-simulation evaluation of tactical-level action plans. Examples are as follows:

[0106] A disaster relief vehicle moving from point A to point B must navigate complex terrain, all of which is affected by rainfall. For example, on an uphill road, if rainfall reaches a certain level, the friction coefficient of the road surface decreases. When the vehicle's load reaches a certain level, it may be impossible to reach the top of the slope.

[0107] Therefore, deciding when to set out (T), when it will rain (RT) and how hard it will rain (RP), how much weight to carry (G), and which path to choose (P) to get to the destination B (L) the fastest while consuming the least fuel (F) is a simple matter of finding the optimal solution for this action.

[0108] {max(L n ),min(F n )}=SFn(T 0,n ,G 0,n ,P 1,n, (RT1,n *RP 1,k ))

[0109] The simulation system described in the present invention will take the values ​​of the above factors within a reasonable range within a reasonable time range, and then substitute them into the simulation entity for simulation, and obtain the result set after traversal and recursion. Therefore, the design idea algorithm of the overall solver described in the present invention is an automatic cyclic recursive solution. Therefore, in the design of the overall solver system, the present invention designs a task scheduler subsystem, which is responsible for traversing the hypothetical parameters and their combinations according to the planning of the "hypothetical database", thereby forming a solution mode for Monte Carlo simulation. Figure 8 As shown in the figure, it is worth noting that the Monte Carlo simulation method requires a high level of computing power. Therefore, when designing the assumption parameters, it is necessary to rationally design key parameters and their value ranges based on the specific task conditions. Proper assumption parameter design can greatly reduce the amount of computation, thereby improving simulation efficiency.

[0110] 2.2.2 Virtual Clock Action Metronome: As mentioned above, the simulation system described in the present invention is not oriented towards the field of real-time simulation, but is a simulation system oriented towards the application scenario of "pre-assessment". Therefore, it is not required that the clock inside the simulation system and the real clock are highly consistent. The clock inside the simulation system can be completely independent of the time in the external real world, and is only responsible for coordinating the synchronization and sequential consistency between the various modules and simulation entities involved in the simulation calculation. Because the simulation modules on which the simulation entity models involved in the simulation calculation rely have very different computational loads, on the same computing power platform, some modules may be easy to calculate the results, but other modules may take a long time to complete the calculation. However, in the real world, the time consumed by these modules to complete the simulation goals is consistent, so within the simulation system, there will be a situation where the simulation task that has been calculated first needs to wait for the completion of the simulation task with complex calculations before entering the next time period.

[0111] For example:

[0112] There are two simulation calculation modules, A and B. A is responsible for calculating the simulation scenario of a vehicle driving on the road; B is responsible for calculating the simulation scenario of radio wave attenuation in three-dimensional terrain. Obviously, B's calculation is much more complex than A's because B may consider many situations, including the impact of different terrains and obstructing objects on radio waves. This may involve a large number of floating-point operations. Therefore, it is obvious that: in the same real-world time of 1 millisecond, the state of the entity simulated by A (such as a car) has hardly changed, while the entity simulated by B (radio waves) may have traveled an extremely long distance and experienced a lot of attenuation.

[0113] In summary, the simulation system described in the present invention has a virtual clock designed inside, such as Figure 9 As shown in the figure, its primary function is to coordinate the various simulation modules to complete tasks that would take the same amount of time in the real world within the same time interval. Within the entire simulation system, this virtual clock is effectively a global metronome, acting like a band conductor. Each beat of the virtual clock's metronome notifies all simulation modules within the system. Each simulation module determines the amount of work it can complete within the time interval set by the system, thereby maintaining synchronization.

[0114] For example, a virtual clock metronome has one beat per minute, a car simulation module will make the car drive one kilometer on the map, while a human walking simulation module will only make the person walk one hundred meters on the map. Of course, this is just a very simple example for ease of understanding.

[0115] The step size of each beat of the virtual clock's metronome is a crucial component of simulation scenario setup. This step size is granularized to varying degrees for different tasks. Generally speaking, the more in-depth the simulation, the more detailed the step size, but this also increases the computational effort. For relatively large-scale simulations, the virtual clock's metronome's step size can even be set to hours or days.

[0116] The simulated virtual clock metronome designed by the present invention can be divided into different levels to meet different simulation fine-grained requirements, but the beat interval of the upper virtual clock (larger time scale) must be the least common multiple of the beat interval of the lower virtual clock.

[0117] 2.2.3 Interaction model based on AI perception: Using AI (Artificial Intelligence) in the architecture of the tactical simulation system is an important innovation of the present invention. The present invention uses artificial intelligence to simulate the interaction at the entity perception level. The main purpose is to reduce the coupling between the various simulation modules and make them reusable. Minimize the interface development cost of the tactical simulation system. The advantages brought by this innovation have been described in detail in the previous article, so I will not repeat them here.

[0118] 2.2.3.1 Overall Design of the Perceptual Interaction Framework: Based on the design concept of "perceptual natural interaction," this invention innovatively proposes a "requested program module." As previously mentioned, the tactical-level simulation system described in this invention utilizes a precise three-dimensional geographic information system (3D GIS) as its digital sandbox foundation. Therefore, all simulation components (simulation units) exchange data through the simulation environment's digital base. This data exchange is divided into two parts: the first is data exchange between simulation modules; the second is data exchange between the simulation entity and the simulation environment. To reduce the complexity of data exchange between simulation entity modules, this invention maps most of the simulated entity's perception of surrounding changes into environmental perception. For example, a tank detects another tank by visually scanning its "field of view" from its position and detecting the target through visual recognition in the image. This simulation approach closely resembles a real-world scenario. This content has been described previously and will not be repeated here.

[0119] The present invention uses a "on-demand interaction model" to achieve this function. Here, the model is explained using visual interaction as an example.

[0120] A simulated entity A needs to obtain a static or dynamic image of its field of view—the entire scene that A visually "sees," referred to here as a ViewImage. It first submits a "on-demand" request to the virtual environment simulation system's digital electronic sandbox system. This request, consisting of an object model (Request Object Module), contains A's precise location and other necessary information (such as the field of view). Based on this object model information, the virtual environment simulation system (digital electronic sandbox), which hosts the environment model, performs a series of temporal, spatial, and environmental calculations to generate a ViewImage starting at that coordinate point. This image is then transmitted to entity A as a program for playback. With this single "on-demand" request, entity A obtains a ViewImage of its spatial and temporal coordinates (X / Y / Z / T). Whether it is a video or image depends on the requirements: some simulation scenarios may suffice with a static image, while others require dynamic video. Simulated entity A can then use machine vision to perform AI recognition on this "field of view" to detect specific targets. Figure 10 It is a schematic diagram of the on-demand interaction model and visual field image.

[0121] A few points to note are as follows:

[0122] First, the environment model hosted by the virtual environment simulation system (digital electronic sandbox, sometimes also called a digital battlefield) is intelligent, capable of intelligently changing the field of view based on the simulation process. For example, the field of view generated by daytime and nighttime changes depending on the time of an on-demand request. Furthermore, different seasons, lighting conditions, and climate conditions all affect the field of view. The environmental model intelligently accounts for these "interference factors." The same applies to non-visual perception. Second, the on-demand interaction model is implemented in information technology through a "Rich Message Queue." This means that all request and response messages are handled through a highly reliable message queue. This ensures that each message has its own timestamp, ensuring that no event occurs at any microscopic point in time, causing the order of events to be disrupted. This is a requirement for any simulation system. Third, because the system incorporates the aforementioned global virtual clock metronome, which provides "step-by-step" rhythm control for all entities participating in the simulation, and each entity model in the simulation clearly understands what it can do within a given time interval, the entire system maintains overall synchronization. Fourth, in special cases, when the on-demand programs generated by the electronic sandbox of the environmental system exceed the carrying capacity of message transmission, hyperlinks can also be used for information transmission. After receiving the message, the simulation entity obtains the required content through the hyperlink.

[0123] As can be seen, this simple yet ingenious architecture minimizes the complexity of "custom protocols"-based interactions between complex models. It allows simulated entities, at both the perception and interaction levels, to group all their perceptions of the external world into a limited number of types, rather than requiring specific data structures for each specific type of perception. This significantly reduces the complexity of system development while increasing its flexibility. In effect, it fully decouples the interactions between simulated entities.

[0124] 2.2.3.2 Types of Perceptual Interactions: This section uses the human body as an example. As a complex system, the human body's external perception channels are primarily the five senses and the skin. These organs handle almost all of the channels through which the human body perceives the external world. New technologies resulting from scientific and technological development are essentially extensions of human perception systems intended to extend and enrich them. For example, telescopes and radars are extensions of vision; walkie-talkies are extensions of hearing, and so on. By using natural interaction to normalize the interactions of simulated entities in a simulation system, the vast majority of perceptual interaction models that can be used in the simulation system can be summarized into a limited number of perception types.

[0125] It should be noted that different technical solutions have different approaches to classifying sensor types. Similarly, the sensor classification used in this invention is also based on the specific technical framework. The basic principle is that technical implementations of the same type will be classified into the same category.

[0126] (1) Visual perception (thermal sensor)

[0127] In the present invention, visual sensors can be roughly divided into sensors in the visible light band and sensors in the non-visible light band. The former includes various cameras, telescopes, etc.; the latter includes non-visible light spectrum imaging devices such as thermal imagers and infrared sensors. The common feature of this type of visual sensors is that as an extension of the visual function of the human eye, the final image still requires the human eye to make a judgment. The interaction at the perception level generated by this type of visual sensor is ultimately submitted in the form of an image (picture or video) to a simulation entity module with artificial intelligence machine vision recognition capabilities for recognition. It’s just that the images generated in different bands are different. As mentioned above, all of these visual field images are generated by a virtual environment simulation system (digital electronic sandbox) that carries the environment model.

[0128] Let's use an infrared sensor as an example to illustrate the generation of field of view images in the non-visible light range. If you need to generate environmental images such as thermal imaging, you need to specify the required temperature range during simulation scenario settings or "on-demand interaction." This is as follows:

[0129] Near infrared region: 770nm~1.5μm Mid infrared region: 1.5~6μm Far infrared region: 6~40μm Extreme far infrared region: 40~1000μm

[0130] Infrared radiation is essentially a form of thermal radiation. Any object with a temperature above absolute zero will radiate energy into outer space as infrared light. As a form of electromagnetic wave, infrared light propagates linearly through space and exhibits common electromagnetic properties, such as reflection, refraction, scattering, interference, and absorption. This requires accounting for the thermal effects of objects and the penetrating power of infrared light. Infrared detectors can be broadly categorized by their principle: thermal detectors and photon detectors.

[0131] The generation of field of view images in specific infrared bands (including other non-visible light bands) is achieved using specific simulation module algorithms. These algorithms are not part of the present invention and are independent generation modules developed by professionals. The output of these modules is a visual image that meets the corresponding optical characteristics.

[0132] (2) Auditory perception (sonar sensor)

[0133] Acoustic sensors primarily refer to microphones (air waves) or hydrophones (liquids) that can capture signals and distinguish sound characteristics such as pitch, timbre, harmony, loudness, and rhythm. Common sensors can distinguish the sounds of motors and engines in different mechanized combat units. Similarly, many vehicles require a certain amount of engine power to support combat operations. Therefore, no matter how quietly or muffled the engine is, the sound of the running engine can be heard. In this invention, the interaction of auditory sensors also uses an on-demand method to transmit sound segments through message sequences. Because sound propagation is far slower than the speed of light, it may not be possible to obtain a complete sound segment within a single virtual clock cycle. Several virtual clock cycles are required to "assemble" a recognizable sound sequence. This is also easy to understand: in fact, when humans understand the sounds they hear, they also listen and analyze them one segment at a time. Sound generation is similar to the process of generating visual field images mentioned above, and is also generated by the environment model system. This system generates different sound effects based on the spatial position of the virtual environment simulation system (digital electronic sandbox). For example, the farther away from the sound source, the greater the added attenuation, making the sound more difficult to distinguish for the simulated entity. If multiple sound sources are emitting sound at a given point on the sandbox, the environmental modeling system automatically blends these sources to create a mixed sound that best simulates a realistic sound field. These sound field characteristics are discretely meshed within the system and dynamically updated as parameters within the properties of each mesh.

[0134] (3) Tactile perception

[0135] Regarding the simulation process of tactical-level actions, especially the process of "mobile modeling and simulation", most entities of the simulation targets need to interact with the environment. For example: different landforms and terrains will have different effects on the forward movement of vehicles; the load-bearing properties of bridges will also hinder the forward movement of vehicles with different loads, etc. This type of interactive perception with the environment is collectively classified as "tactile" perception in the present invention. Tactile perception is still applicable to the aforementioned "on-demand interaction model". When the simulation entity needs to perceive the surrounding environment that can be touched, the environmental model system is called with grid parameter coordinates to obtain information about its location. Unlike visual and auditory perception, tactile perception returns an extensible object result model, which includes the environmental parameters, terrain, landform parameters and other custom parameters required for the simulation task. At the same time, the content returned by the object model also includes the methods (program code) required for the calculation of certain environmental parameters. Described in object-oriented language, the object model includes the attributes of things and the corresponding calculation methods.

[0136] It's important to note that at any three-dimensional grid coordinate location (X, Y, Z, T), different timestamps present different properties. This is easy to understand because the simulation environment, which aims to "simulate the real environment as closely as possible," is dynamic. Therefore, the content of environmental model feedback includes both static and dynamic components. The environmental model virtual environment simulation system (digital electronic sandbox) can generate three types of "tactile" feedback:

[0137] First, static environmental parameters are generated based on the scenario settings. Examples include terrain slope and the density of surface vegetation according to the set season. These parameters can be considered "static" in a tactical simulation. Second, environmental parameters may change continuously as the simulation progresses through virtual time. For example, road friction may vary with precipitation levels. Third, interactive feedback refers to the impact of the simulation process on the environment. For example, road bridges may be damaged by artillery fire, reducing or even rendering them impassable. Similarly, smoke from burning trees may reduce visibility. Environmental feedback is returned to the object model via a message system. This object model requires object parsing by the simulation entity module program. The object model is uniformly defined by the environment model system, but uses a common data structure standard. The virtual environment simulation system (digital electronic sandbox) that hosts the environment model uses a data organization format called "grid objects," which will be described later.

[0138] (4) Electromagnetic sensing (radar, electronic reconnaissance-electronic support measures, electronic jamming measures)

[0139] Radar and sonar systems play similar roles in sensing and both play an important role. They are both used to search, detect, and track targets, as well as guide weapon systems into target areas. The difference lies in that radar uses electromagnetic waves that propagate through the atmosphere, while sonar systems use sound waves that propagate through water. Unlike sonar systems, which passively receive sound signals, radar operates by actively generating electromagnetic waves, reflecting them off the target, and then receiving the reflected signal. Its detection is essentially based on the evaluation of reflections. A typical differentiating factor between radar systems is the waveform used: pulse radar uses spatial differences (reflection time, target position), Doppler radar uses the Doppler effect (frequency modulation, target velocity), and frequency modulated continuous wave (FMCW) radar uses both (position and velocity). The operating principle of radar can actually be precisely described using mathematical models, which is not the focus of this article, but the most basic part is used to calculate the signal-to-noise ratio.

[0140] Because this system utilizes a precise 3D GIS model based on the actual Earth's curvature as its simulation digital chassis, the mathematical calculations for specific radio wave propagation on the Earth's surface and in the atmosphere are packaged into a separate module designed by professionals. In other words, once the terrain for the operation is determined, the specific behavior of radio waves on this fixed terrain can be calculated using a precise mathematical model. Therefore, within the entire 3D grid system, the signal strength of radio waves emitted from the same transmission origin in each grid cell can be calculated. By factoring in other interference factors such as weather, the precise value of the radio signal at each grid cell at a specific time is determined. When a simulated entity needs to receive a radio signal on a specific channel, it simply sends its current coordinates and time to the environmental model system to receive feedback. This feedback even includes any interference information already incorporated into the signal. Therefore, the simulation system described in this invention only needs to obtain the results of the radio signal superposition from the environmental model system. However, it should be noted that different radio equipment utilizes radio waves in different ways, and each type of radio equipment is responsible for its own mathematical model, which is the responsibility of the professional radio modeling team. However, the interference items superimposed in the actual environment will interfere according to their own calculation methods. Figure 13 This is an event information flow diagram. For example, consider the case where a signal transmitted by one radio station is received by another. In a real-world environmental model system, key factors such as the frequency band used, transmit power, antenna type, and modulation method (calculated by professionals with extensive knowledge of the equipment) are ultimately combined with factors such as topography and climate, and transmitted to the receiver via a specific object model. Because the grid through which the radio waves pass has a pre-calculated signal-to-noise ratio for a specific time and space, the signal strength received by the receiver in a straight-line state is weakened by this interference, which ultimately directly impacts the performance.

[0141] 2.2.4 Object-Oriented Container Encapsulation Technology

[0142] Object-oriented thinking is a remarkable invention of human thought. In fact, objects are tools for modeling real-world objects. Object-oriented technology has been applied to simulation modeling for a long time, as people have long recognized the ability to describe real-world objects as objects. Traditional simulation systems also employ object-oriented technology, such as the MDA mentioned above. However, in computer technology, object-oriented development techniques are often tied to specific development tools and languages. Consequently, models developed using different development tools require a complex set of protocols for interaction. For example, the US military's SEDRIS standard (Synthetic Environmental Data Representation and Exchange Specification) contains over 1,600 data representation specifications, enabling various military applications to operate on the same Master Environment Database (MEL). This standard establishes a complex set of interfaces, requiring systems developed using different software development tools to operate within this unified database according to these fixed, standard interfaces. All of these issues arose under the current information technology conditions. Under new conditions, with the emergence of virtualization and containerization technologies, simulation modules developed using different development languages ​​or tools can be encapsulated into a unified container. Even a simulation module can be housed in different containers.

[0143] This invention utilizes container platform technology to describe and encapsulate a single simulation entity or part of a simulation entity. This technology leverages and combines containerization and traditional object-oriented technology to form an object-oriented container platform. Object-oriented technology has become mainstream today and is essentially an upgrade to object-oriented technology. A container can be considered a more complex "macro-object." The key difference between this and traditional object-oriented technology is that each "object" can be developed using different development tools. This allows specialized simulation entity algorithm modeling to be independently completed by specialized teams. These teams can then collaborate using the object-oriented container platform. For example, specialized content like the mathematical model of terrain effects on radio wave propagation mentioned above could be developed using any development tool, but now it is encapsulated with a unified interface. Containers encapsulate both data and methods. The underlying layer of the object-oriented container platform is a business-independent container orchestration and scheduling system, implemented using the industry's most common Kubernetes architecture. The containers themselves encapsulate not only simulation entities but also the aforementioned MQ message entities (invoked by passing the container address), as well as other simulation modeling models and scenario parameters. For example, in the environmental model system mentioned above, the signal attenuation model calculation involving the grid is implemented in the form of object containers.

[0144] 2.2.5 Simulation modeling technology

[0145] Simulation modeling technology simulates the behavior and performance of real systems by building system models. It can be used to study how a system operates under different conditions. For example, planning a contingency plan for transporting disaster relief supplies during a disaster requires determining the maximum transport volume and minimum transport time under varying weather conditions. This requires modeling and simulation of terrain, weather, vehicles, and other relevant factors.

[0146] The simulation system constructed by this invention primarily simulates tactical-level operations under different conditions. For example, different time points and weather changes during an operation can influence the success of the operation. This requires two broad categories of simulation for this type of simulation: environmental simulation and action unit (or combat unit) simulation.

[0147] 2.2.5.1 Accurate Geographic Information System Interactive Computing and Dynamic Environmental Modeling Technology

[0148] As previously described, environmental simulation primarily refers to a simulation system built on a precise three-dimensional geographic information system (GIS) and overlaid with models of key environmental variables. The former primarily refers to static geographic environments such as topography, while the latter refers to dynamic natural environments such as wind, frost, rain, and snow. Because of the high-precision modeling of a realistic three-dimensional battlefield environment, the entire battlefield space can be easily divided into grids of varying sizes, with each grid being considered a "cube." Depending on the required simulation accuracy (granularity), the scale of these "cubes" can also be set to different sizes during modeling scenarios. For dynamic natural environmental factors such as weather conditions, separate mathematical models are established for each dynamic weather factor. By setting different parameters during the environmental scenario, the continuous functions of these mathematical models are discretized, with the specific granularity determined by the grid size. This process is generally referred to as "discrete sampling." Thus, for the mathematical model of the same dynamic factor, each grid cube will reflect the value of the continuous function of that dynamic factor, with the coordinates of the grid as parameters. The environmental simulation system supports dynamic feedback. The feedback of the environment is returned to the object model through the message system. The object model requires the simulation entity module program to perform object parsing. The object model needs to be uniformly defined by the environmental model system, but it adopts a general data structure standard. The virtual environment simulation system that carries the environmental model uses a data organization form called grid object.

[0149] In summary, the modeling process is essentially solving differential equations for the natural phenomenon. The simulation process, on the other hand, typically involves solving partial differential equations for a specific situation, then discretely sampling the results and assigning them to the environmental grid system.

[0150]

[0151] The above formula is a mathematical model of radio wave propagation in free space, which includes parameters such as the received power at the distance from the transmitter, the radiated power of the antenna, the gain of the transmitting antenna, the gain of the receiving antenna, the system loss factor independent of propagation, the wavelength and the distance between the transmitter and the receiver.

[0152] In other more rigorous simulation tasks, other factors need to be considered, and other models need to be used. For example, the two-path model of radio wave attenuation:

[0153]

[0154] Let's take an example: We model a rainfall process, forming a rainfall function. The main parameters of this rainfall function may include cloud thickness, cloud coverage, and other important weather factors that may affect rainfall. This can be a rather complex mathematical model, but this is not the focus of this invention. Typically, such mathematical models are developed by professional meteorological researchers. When incorporated into the simulation system described in this invention, the model code is already developed, mature and complete, and encapsulated into an object container. The simulation scenario system obtains different results by setting different parameters for this container. The calculated results are then copied into the aforementioned grid system. Ultimately, each grid has discrete sample values ​​(in this case, rainfall) that conform to the geographic coordinate system for the mathematical model. As simulated entities move in a three-dimensional environment, they will sequentially pass through these specific cubic grids along their movement routes. Simply reporting the current coordinates and time of the moving entity allows us to obtain rainfall values ​​of varying magnitudes from the current grid. Each grid is effectively an object and is extensible. They can be redefined to meet the requirements of different simulation applications.

[0155] It can be easily seen that similar examples to the above, including wind, light, terrain, radio wave transmission, sound transmission, etc., can all be handled in the same way.

[0156] Therefore, in the process of modeling the environmental simulation model, the solver architecture design scheme described in the present invention can clearly decouple algorithm models of different complexities from specific simulation tasks, so that these environmental algorithm models can be reused, and integrated with the static geographic information environment system through the object model to form a joint effect on the active simulation entity. It is worth mentioning that the above-mentioned natural environment will also be affected by the action unit. For example, a bridge may be broken if it is hit by artillery fire. This situation has a huge impact on the actual simulation task and should also be taken into account. This situation is called "dynamic environment interaction." The tactical-level simulation system and solver described in the present invention have the ability to interact with the dynamic environment.

[0157] 2.2.5.2 Action Unit Simulation

[0158] Action (combat) units are entities, such as vehicles, personnel, aircraft, and ships, that participate in tactical-level action mission simulations, independent of the battlefield environment. These entities are not limited to movable entities; they also include immovable entities, such as radar stations. These action units carry out planned operations within the aforementioned environmental simulation system, interacting with the environment and being both influenced by and impacting it.

[0159] Simulation of a mobile entity primarily encompasses two aspects: first, the simulation of its own properties; and second, the simulation of its passive behavior under environmental influences. For example, taking a car moving on a road (often referred to as "mobility simulation"), the former involves modeling the power, load capacity, braking capacity, and other static and dynamic properties of a specific vehicle model, and providing key parameters influencing these capabilities (such as the type of fuel used). The latter simulates how these physical characteristics change under environmental influences. For example, at high altitudes, the low oxygen content affects engine combustion, resulting in a decrease in vehicle power. This can lead to a slope that was previously climbable on plains becoming impossible in plateaus due to insufficient power, thus impacting overall mobility. Furthermore, when vehicles of varying loads cross bridges, the load-bearing capacity of the bridge must be considered, making route selection particularly crucial. Heavy rain, slippery roads, and other factors can also affect vehicle movement, among other factors. This encompasses the behavior of vehicles (mobile entities) under environmental influences.

[0160] In the present invention, the simulation method for action entities is essentially the same as the simulation of the aforementioned environmental dynamic factors. Because certain specialized action entities require physical modeling by professionals, for example, only engineers at automobile manufacturers have the most profound understanding of the vehicles they produce. The present invention employs an object container model solution, allowing the modeling of these specific action entities to be performed by organizations that understand them best (e.g., equipment manufacturers), and they can use any development tools for modeling. The object model composed of the completed model can include both the physical parameters of the target model and the specific behavioral responses exhibited by the model in the face of external dynamic changes. For example, different air oxygen contents (all stored in the object model of the aforementioned "environmental grid system" and implemented through "environmental interaction") affect engine combustion and even power.

[0161] Modeling and simulating other types of entities are basically consistent with the above principles and belong to "entity-level simulation". However, what needs special treatment in the solution process is the "aggregate-level simulation" of entities. The former belongs to the category of individual tactical level, but if the simulation task requires a more macroscopic or fine-grained level, the latter aggregation-level simulation needs to be used. For example: a car squadron consisting of 10 cars, this is an aggregation-level simulation. If all the entity types in the aggregate object are the same, such as a convoy of 10 cars of the same type, this is called "homogeneous aggregation". If the entity types in a group are different, such as a convoy of 10 cars of different types, this is "heterogeneous aggregation". Figure 11 It is a schematic diagram of the interaction between the simulation entity and the environmental grid system.

[0162] The present invention uses the object-oriented runtime class (Classes at Runtime) method to handle the situation of aggregate simulation modeling. First, the basic independent entities are modeled, and then the basic object container is constructed. Then, during the simulation run, based on the settings of the assumed system, a mechanism similar to "Reflection" in object-oriented programming is used to create runtime classes and perform aggregate expansion. In other words, the class of a single entity is inherited as needed at runtime to become an "aggregate entity" with the same properties. This solution simplifies the development required for aggregate-type simulation and provides great flexibility.

[0163] For homogeneous aggregation, if the simulation resolution requirement is not high, we can simplify it into one entity for simulation to save computational effort. For heterogeneous aggregation, it is necessary to consider four aspects: model construction, data processing, interaction management, and scheduling strategy. The present invention adopts a classification processing solution to convert heterogeneous aggregation into homogeneous aggregation entities or even independent entities through decomposition. After decomposition, the number of aggregation entities is less than or equal to that of independent entities. The decomposed homogeneous aggregation entity becomes an independent class, which is instantiated and encapsulated in an object container when the solver is run.

[0164] 3.3 Engineering Implementation Plan

[0165] The engineering implementation of the present invention relies on the development of new information processing technologies. This section describes the engineering implementation of the key points of the implementation scheme described in the previous section.

[0166] 3.3.1 Overall Project Implementation Plan

[0167] Generally speaking, the system described in the present invention uses the following design concept in its engineering implementation, namely: fully decoupling the simulation process through full objectification, containerization and standard interface methods. Under the guidance of this overall concept, the types of components and interfaces are reduced as much as possible during the system implementation process. According to the design concept of exchanging space for complexity, the entire system forms an architecture with a large number of identical components but a small number of component types. This design form facilitates the realization of mechanized unified management, and the overall system architecture formed is as follows: Figure 12 As shown in the figure, the system is divided into three parts: the client, the service cluster, and the platform layer. The platform layer is business-independent. All other business-related functions are implemented as object containers in the engineering implementation, with consistent structure and interfaces. This maximizes standardization and scalability, thereby achieving full decoupling in engineering design.

[0168] 3.3.2 Unified Data Base Solution Based on High-Precision Geographic Information System

[0169] The data base mentioned in this invention is actually a common standard for all data used in the simulation system. Geographic information data, which is most closely related to tactical-level simulation, is selected as a unified reference. A high-precision geographic information system can be used as a "ruler" for all simulation applications in the simulation system. In other words, the relationships between all simulation units are connected through this "ruler," both in time and space.

[0170] In engineering implementation, the present invention "maps" all these simulation units into a high-precision geographic information system, and connects them through specific spatial coordinates and time coordinates. So we call it a "data base". In the present invention, the data base uses a high-precision three-dimensional geographic information system, which can use a plane coordinate system or a spherical coordinate system. When the simulation area is relatively small, a plane coordinate system is used to reduce the amount of calculation. When the simulation area is relatively large, the curvature of the earth's surface can no longer be ignored, and a true spherical three-dimensional coordinate system is required. In order to avoid relying on any existing system in engineering implementation, the present invention can use any high-precision three-dimensional geographic information system as a data base. The main function provided by these geographic information systems is the coordinate system calculation service.

[0171] This invention uses a three-dimensional (3D) grid coordinate system to perform coordinate conversion with the coordinate systems of these high-precision geographic information systems. This task is accomplished by the "Coordinate Conversion Computation Service" module within the system constructed by this invention. This module involves a large number of floating-point operations, representing a significant amount of computational effort during the entire system design process. Therefore, in engineering design, this module was implemented using a dedicated cluster equipped with high-performance floating-point computing. Architecture-wise, a mature, highly resilient computing cluster was employed to achieve high-load computing capabilities. These are mature technologies, not the focus of this invention's briefing document, and will not be elaborated upon here. Simulation tasks have varying degrees of granularity, which in the simulation system is referred to as simulation resolution. Therefore, the resolution of this three-dimensional grid coordinate system is dynamically adjusted as the task changes. However, regardless of how it is adjusted, it maintains a proportional relationship with the actual geographic information data. This not only creates a unified reference system but also significantly reduces computational complexity.

[0172] It's important to note that the high-precision geographic information system mentioned here is not inherently business-related. We simply aim to use it to host the fundamental data for the simulated geographic information environment and, to the greatest extent possible, to link all simulation units and corresponding public simulation services to geographic coordinates. Combined with the aforementioned virtual clock and metronome system within the system, this creates a virtual spacetime. In other words, each simulation unit and its associated components are endowed with spatiotemporal attributes. Only under the same spatiotemporal standard can all simulation units and their component services interact.

[0173] As mentioned earlier in this article, the dynamic nature of GIS (Geographic Information Systems) means that as the simulation progresses, simulation events may cause changes to the spatial data contained in the GIS (virtual environment simulation system – digital electronic sandbox). For example, in the disaster relief simulation mentioned earlier, severe meteorological disasters can cause changes in topography, such as destroying bridges and disrupting traffic. To address this, we implemented a "topography snapshot" solution. This solution saves a snapshot of the spatial data contained in the grid where the change occurred, then applies the corresponding data changes to the same grid for subsequent calculations. This way, as the simulation system's time ticks, the topography (spatial data) at that location dynamically changes with the simulation process. When the simulation is interrupted or restarted, simply restore the snapshot of the spatial data from any previous point in time.

[0174] In summary, in this system's design, the GIS is decoupled from the upper-layer simulation applications. In most scenarios, it functions as a spatial database, providing only coordinate resolution and data return services. However, it also plays a key role in providing a unified reference system throughout the system's operation.

[0175] 3.3.3 Engineering Implementation of Distributed Object Container

[0176] The basic container platform used in this paper is the most popular K8S. K8S stands for Kubernetes. Its features include:

[0177] ●Automated deployment and rollback: Containerized applications can be deployed automatically, and version rollbacks can be performed easily when problems arise. For example, when a new application version fails, it can be quickly restored to the previous stable version. ●Elastic scaling: The number of application replicas can be automatically adjusted according to the load situation. For example, during e-commerce promotions, the number of service replicas can be automatically increased to cope with high traffic. ●Service discovery and load balancing: IP addresses can be automatically assigned to containers, and traffic can be evenly distributed to each container instance through the built-in load balancing mechanism. ●Storage orchestration: Storage systems can be automatically mounted, whether it is local storage, network storage, or cloud storage, to facilitate application data persistence.

[0178] K8S itself is a basic information platform that has nothing to do with simulation business. The present invention uses its container management characteristics to achieve decoupling between modules. However, as mentioned above, the present invention does not rely entirely on the standard system architecture of microservices, but uses the container mechanism to build a "distributed object model" and encapsulate it in the container. Then, use K8S's orchestration and scheduling functions for containers to operate the containers. At the same time, make full use of the highly elastic clustering capabilities of the K8S platform to realize simulation calculations of complex tasks. The principle description of the "distributed object model" has been mentioned above and will not be repeated here.

[0179] This project implementation method draws on the microservice container architecture and leverages the maturity of the Kubernetes platform. It also improves upon the existing microservice model, making it object-oriented. Objectification is the most intuitive way to model real-world objects.

[0180] Another issue that needs to be considered in engineering implementation is the amount of communication interaction between containers. It can be seen from the design of the system that in this case, because the simulation computing unit and the public service part are designed as a microservice model, the amount of communication between each microservice container will become extremely complex. In order to reduce the performance impact brought about by these communications, the communication between each microservice container in the present invention uses a compressed binary communication protocol (the present invention uses Google's Protocol Buffers). This binary-based encoding protocol is twenty to one hundred times faster than the text-based JSON protocol commonly used in microservice architectures. In order to achieve this performance advantage, additional encoding and decoding work is required. Because the original microservice protocols are basically based on JSON, this requires the construction of a business-independent encoding and decoding module. Make full use of modern mature technology to carry the innovative mechanism of the present invention to avoid instability in engineering implementation to the greatest extent.

[0181] 3.3.4 Engineering Implementation of Event-Driven Model

[0182] like Figure 15As shown, the simulation system described in the present invention is like a symphony orchestra with many members. All "band members" must obey the orders of the "conductor", and the notes and strength of each beat must be determined by the conductor's gestures. In this system, this "conductor" is the virtual clock metronome described in the technical principle part b above. It is the driver that drives all the object models participating in the simulation of the entire system. After receiving a specific beat signal, each object model participating in the simulation calculation determines what it can do in this specific time interval based on its own internal calculation mechanism, and then sends the resulting status to the corresponding other simulation objects to form an interaction. And this interaction is synchronous. This is because the causal logical relationship between the various simulation units must be maintained, just like in the real world.

[0183] As described above in the principle of the invention, since the system constructed by the present invention is a "non-real-time" simulation application, all interactions are "synchronous" and do not need to be linked to real-world events, but simply calculate results according to a rhythm. Therefore, the event-driven engineering implementation is normalized to a "message queue" implementation. The system of the present invention uses the mature Rabbit MQ service as the message middleware, and the protocol uses the highly reliable AMQP protocol (Advanced Message Queuing Protocol). AMQP is primarily used for highly reliable message delivery, handling scenarios such as order processing, financial transactions, and military applications in complex enterprise environments. AMQP has a richer message model, supports multiple exchange types such as direct connections, topics, and sectors, and can flexibly handle complex message routing. AMQP provides multiple mechanisms such as transactions and message confirmations to ensure reliable message delivery, including message accuracy and timestamp ordering, ensuring that "first sent" will always "first arrived." These characteristics are precisely what the system must possess in simulation applications. Using AMQP as the messaging protocol maximizes the maturity of existing systems for the engineering implementation of the present invention. Figure 14 This is the MQ cluster and high availability deployment diagram of the present invention.

[0184] RabbitMQ is mature and outstanding in terms of reliability and stability. It is written in Erlang and has high concurrency and distributed features, which can cope with large-scale message processing scenarios. For example, in the scenario of processing financial transaction message queues, it can effectively ensure the reliable delivery of messages. In the present invention, because all messages (encapsulated into object models in the present invention) must be synchronized, the load and reliability requirements of MQ are very high. Therefore, MQ services form elastic clusters (Clusters), which can be expanded as the complexity and performance requirements of simulation tasks increase, realizing a solution of trading resources for time. In fact, there is only one thing that flows between each object model in the entire system - "events" (time beats are also a type of event). Therefore, the simulation system described in the present invention can be understood as a completely event-driven information system. After receiving a specific event, each unit can choose to ignore it or choose to produce a specific result according to customized rules. When the simulation units are performing simulation applications, they also interact with each other according to the corresponding interaction rules through the aforementioned message-based standard protocol.

[0185] 3.4 Application Mode

[0186] The simulation system described in this invention employs a computational simulation mechanism known as Monte Carlo simulation. Therefore, in principle, no human intervention is required during execution. After setting the scenario, the operator typically only needs to wait for the results. The system automatically drives the participating simulation units to operate according to the specified settings and a specific workflow.

[0187] The core application steps of the tactical simulation solver system constructed by the present invention are: Step S1: describe the problem and define variables; Step S2: decompose the problem and build a model; Step S3: generate conditional parameter samples, that is, set the assumed conditions; Step S4: simulation calculation; Step S5: simulation effect statistics; Step S6: result confidence assessment.

[0188] 3.4.1 Deployment Plan Overview

[0189] The system deployment scheme constructed by the present invention is as follows Figure 15As shown. · Hardware Layer: A cluster of hyper-converged servers (HCIs) is used. This type of equipment offers excellent scalability and a single device type, enabling linear expansion or reduction of cluster size. · Hyper-converged Software and Operating System Layer: The server cluster itself comprises a hyper-converged layer and a virtual machine layer on top of bare metal hardware. The virtual machine layer is installed with a standard operating system and can share the computing and storage resources of the hyper-converged cluster. It also offers excellent fault tolerance and high availability. · Container Platform Layer: A container platform cluster (K8S) is deployed across multiple virtual machines, primarily hosting containers for different purposes. · Basic Service Layer: Containers host basic services, including a virtual environment simulation system (digital electronic sandbox), a high-precision geographic information system, a message queue center, a virtual clock and metronome system, an event-driven center, and various common basic services. · Simulation Application Layer: Various simulation units involved in the simulation and their corresponding algorithms (including environmental simulation algorithms and AI-related models) are encapsulated in containers and deployed on the container platform through its interface. Each container has its own independent configuration environment file for the system. By configuring the configuration environment file, all containers can communicate with each other within the system platform and establish a communication mechanism. When the simulation units or algorithm components involved in the simulation calculation are updated, only the container image needs to be uploaded and updated, and the old version can be restored. Install the corresponding client software on the local or remote terminal, connect to the unified service gateway of the simulation platform through the configuration of the client software, and perform application-level configuration and operation on the simulation system on the client software.

[0190] The present invention also discloses a tactical simulation solver system, including a client, a service layer and an object container orchestration and scheduling platform layer, wherein the service layer includes a simulation scenario configuration service, a clock metronome service, a high-precision 3DGIS platform service, a general AI perception module, a playback rendering service, an interactive object model, a simulation effect object model, an environmental object model, an aggregate entity object model, and a simulation entity object model. The simulation scenario configuration service is an interface (UI) for users to describe specific simulation tasks. Through the visual operation of the software, the following settings are made: a virtual clock, environmental data base parameters (including static parameters and dynamic parameters), operating rules of each simulation unit module, simulation condition screening, effect parameters, and result validity screening rules, etc. The system converts the data input by the user and puts it into the scenario database. The Clock and Metronome Service coordinates simulation modules to complete tasks that would be completed in the same real-world time interval within the same time interval. It can be divided into different layers to meet different simulation granularity requirements. However, the tick interval of the upper-level virtual clock is the least common multiple of the tick intervals of the lower-level virtual clocks. The High-Precision 3DGIS Platform Service builds a unified data base to achieve spatial-temporal correlation between simulation units. The high-precision 3D GIS uses a planar coordinate system and / or a spherical coordinate system. When the simulation area is small, a planar coordinate system is used, while when the simulation area is large, a true spherical 3D coordinate system is used. General AI Perception uses artificial intelligence to interact with simulation entities at the perception level, reducing coupling between simulation modules and ensuring reusability. The Replay Rendering Service visualizes the current simulation status as the simulation progresses, creating a 2D or 3D situation map and placing the playable 3D rendering results in a first-in, first-out (FIFO) queue. This service is controlled by the virtual clock and metronome and AMQP's highly reliable ordered message queue to faithfully reproduce the simulation process. The rendering process is asynchronous with the simulation progress, aiming to minimize the impact of simulation visualization on computing power. Simulation results can be viewed with a limited time delay or replayed after the simulation completes. The interactive object model is a complex data structure used for interaction between different simulation objects and between simulation objects and the environment model. This structure conforms to all the elements of an object, including construction, destruction, parameters, methods, and other sub-objects. The system communicates complex information through this semi-structured object model. The simulation entity object model is a method for modeling simulation objects using an object-oriented approach. Each simulation object represents real-world transactions through one or more objects. This modeling approach leverages the characteristics of object-oriented programming (OO), intuitively describing the real world and facilitating the formalization of relationships. For example, "team relationships" between members can be implemented through object aggregation, while improved equipment can be represented through object inheritance.And through containerization, different objects can be developed using different development tools, which greatly reduces the difficulty of development and supports multi-team collaboration. Environmental object model: Modeling of different elements in the environmental model subsystem. Each model is expressed in an object-oriented way, and these objects include the algorithmic response of the element to different external inputs. For example, the object model of the mud ground and the object model of the grass ground have different friction values ​​given by the vehicle pressure parameters when the tires run over them. Simulation effect object model: An object structure model used to describe the simulation effect using an object-oriented approach. It is used to model the results of the interaction between the simulation entity object model and the environmental model. For example, the smoke generated by an explosion, and this smoke may spread in the environmental model according to a certain algorithm with the wind (a type of environmental parameter). Aggregate entity object model: An aggregate model structure that includes multiple sub-simulation object models. This structure is used to describe a combination of multiple isomorphic or heterogeneous members. For example: a fleet composed of different types of vehicles; object container orchestration and scheduling platform layer: based on K8S or similar container orchestration and scheduling architecture, it uses its container management characteristics to achieve decoupling between modules, uses the container mechanism to build a distributed object model and encapsulates it in the container, uses K8S's container orchestration and scheduling functions to operate the container, and uses K8S's highly elastic cluster capabilities to achieve simulation calculations of complex tasks.

[0191] 3.4.2 Usage Process and Configuration

[0192] The system constructed by this invention minimizes the coupling between components, particularly by integrating heterogeneous algorithms or software developed by different manufacturers or research institutions in a standardized manner. Therefore, it adopts a strategy of "trading time for complexity" in its use. The specific steps for performing simulations using the system constructed by this invention are roughly divided into: Step a1: System Preparation; Step a2: Model Data Preparation; Step a3: Assumption Conditions and Data Entry; Step a4: System Simulation Operation; Step a5: Results Data Analysis; Step a6: Comprehensive Solution Research; and Step a7: Compiling a Summary Report.

[0193] The system constructed using this invention requires the following seats (or user roles), each role or seat is composed of 1 to N people. The specific configuration and definition are described as follows:

[0194] (1) Comprehensive analysis seat: mainly responsible for system definition and preparation, and overall command and coordination of the entire simulation work; (2) Assumption entry seat: decompose the simulation task, form the corresponding assumption conditions and enter them into the system; (3) Operation control seat: always observe the operation status of the simulation system and whether the corresponding intermediate results meet the requirements of the simulation task during the simulation operation, and deal with any abnormalities in a timely manner; (4) Situation display seat: the visual output control display seat for the simulation process and results; (5) Online statistics seat: during the operation of the simulation task, this seat performs statistical analysis and reports on various data; (6) Evaluation and analysis seat: performs comprehensive evaluation and analysis on the simulation results, determines the confidence level, and forms results and reports.

[0195] All of the above seats have their own terminal systems, and the number of terminals can be deployed as one or N, depending on the complexity of the simulation task.

[0196] Innovations of the present invention:

[0197] 1. Apply deterministic conditional Monte Carlo simulation scheme to tactical level action simulation.

[0198] This is a method and solver implementation invented by taking advantage of modern highly developed computing power to perform traversal calculations on a range of deterministic condition sets, thereby obtaining N optimal solutions under different combinations of multiple conditions. The main application scenarios include tactical-level decision support and simulation analysis and evaluation of emergency plans, and it has broad application prospects and value.

[0199] 2. Decoupling model of action simulation unit of "natural interaction technology" based on artificial intelligence perception.

[0200] The "natural interaction technology" proposed in this paper is designed to simplify the interaction between simulation units, reduce system coupling, and improve the reusability of unit models. This technology is inspired by the idea that humans perceive the world through only five primary means: sight, hearing, smell, touch, and taste. The simulation unit itself is an AI agent, ultimately approaching human intelligence. Therefore, using a natural interaction perception solution instead of the software data interfaces commonly used in traditional computer systems not only simplifies system design but also brings it closer to the real world, making the simulation process infinitely closer to real-world intelligent interactions.

[0201] 3. Key points in the engineering design of the tactical-level action simulation solver system:

[0202] a) Distributed object model and containerized encapsulation of the object model.

[0203] Objects are the best technical approach for modeling real-world objects. This invention creates a data structure for a distributed object model, which is used to describe everything, including components and events, in each simulation unit of a simulation system. Using a distributed object model allows for one-to-one modeling of nearly all real-world objects, and provides inheritance and generalization capabilities for objects of the same type. Objectifying all objects in a simulation system allows for more logical relationships between all elements in the entire simulation system (including algorithms, data, and other dynamic information), while greatly improving the reusability of modules.

[0204] b) Virtual clock metronome system and event-driven model.

[0205] In order to simplify the interaction complexity of the simulation unit objects in the simulation system described in the present invention (because the types of such simulation unit objects cannot be enumerated and the number is extremely large), and at the same time ensure continuous strict logic in the simulation application, the present invention uses events to express all the "activities" and "interactions" in the system. And under the command of a unified clock metronome system, a "synchronous interaction" method is used to ensure the seriousness of the logic. In the event-driven model, what is transmitted is also an object model, so that it can express complex information, and even the method of processing information is included. This event-driven model ensures simplicity and efficiency while taking into account flexibility.

[0206] c) On-demand interactive model technology:

[0207] The on-demand interactive model is a specialized method for interacting between simulation units and a virtual environment simulation system (digital electronic sandbox) built on a high-precision geographic information system. This novel approach solves the interaction issues between simulation units and their virtual environment and simplifies and standardizes simulation system development.

[0208] d) Interactive high-precision geographic information system technology:

[0209] This key innovation overcomes the drawback of traditional geographic information systems that can only display static data. It enables the simulation of the interaction between simulation units and the environment in the simulation system, reflecting the dynamic nature of the real world more realistically.

[0210] e) Using "asynchronous visual playback" technology:

[0211] The advantage of this key innovation is that fewer computing resources (computing power resources) can be fully invested in the simulation process, without slowing down the simulation process due to the large amount of computing resources occupied by the visualization display. The effects produced during the simulation process are first placed in a visualization buffer in the form of executed command scripts. This buffer has a first-in-first-out (FIFO) queue structure, and the device responsible for visualization rendering reads the corresponding visualization instructions from the exit of this buffer. The visualization rendering function module executes the visualization instructions at its own pace to produce the visualization output. Because of the existence of the buffer, even if some simulation processes take too much time, it will not affect the continuity of the visualization output.

[0212] The above is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. For those skilled in the art of the present invention, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should be considered to fall within the scope of protection of the present invention.

Claims

1. A method for simulating joint actions using a deterministic conditional Monte Carlo simulation scheme, characterized in that: include: Monte Carlo simulation steps: Using the concept of deterministic conditional Monte Carlo simulation, a traversal calculation of key element combinations is performed for the action target to be simulated. Utilizing the high-speed computing power of the computer system, all combinations of key parameters that meet the task requirements are input into the simulation entity for calculation. Ultimately, n results are automatically obtained, and then m optimal results are screened out from them. Virtual clock metronome step: Design a virtual clock metronome to coordinate various simulation modules, decouple it from real time, and use it to synchronize and coordinate simulation modules to complete tasks required in the real world within the same time interval; Perception interaction step: Use artificial intelligence to interact with the simulation entity perception layer to reduce the coupling between various simulation modules and make them reusable; Object-oriented container platform steps: Use container platform technology to describe and encapsulate a single simulation entity or part of a simulation entity, fully utilizing and combining containerization encapsulation technology and traditional object-oriented technology to form an object-oriented container platform; Modeling steps: Build a tactical-level action simulation system, use environmental simulation modeling as a digital base, and combine it with action unit simulation modeling to simulate the impact of different conditions on action results.

2. The joint action simulation method according to claim 1, characterized in that: In the virtual clock metronome step, every time the virtual clock metronome beats a beat, it will notify all simulation modules in the system. Each simulation module determines the work it can complete within this time interval based on the time interval of each beat set by the system, and thus maintains synchronization; in the virtual clock metronome step, when the virtual clock metronome is divided into different levels to meet different simulation fine-grained requirements, the beat interval of the upper-level virtual clock is the least common multiple of the beat interval of the lower-level virtual clock.

3. The joint action simulation method according to claim 1, characterized in that: The perceptual interaction step further includes: On-demand interaction model steps: The on-demand interaction model is used to optimize the data interaction between simulation entities; all simulation components exchange data through the simulation environment digital base. The data exchange here is divided into two parts: the first part is the data exchange between simulation modules; the second part is the data exchange between the simulation subject and the simulation environment; in order to reduce the complexity of data exchange between simulation entity modules, the simulation entity's perception of surrounding changes is mapped into the perception of the environment. Through the perception method of the environment intermediary, the simulation logic is ensured to be infinitely close to the real scene, thereby enhancing the realism of the system.

4. The joint action simulation method according to claim 3, characterized in that: The on-demand interaction model steps include: Visual interaction model steps: Simulated entity A first submits an on-demand request to the virtual environment simulation system. The request is an object model that contains the precise location and other information of simulated entity A. The virtual environment simulation system that carries the environment model will calculate time, space and environment based on the information of this object model, and generate a field of view image starting from the coordinate point, and then pass it to simulated entity A as a program for playback. In this way, simulated entity A obtains the field of view image of its time and space coordinate point (X / Y / Z / T) through a single on-demand operation. Specifically, it is a video or image, which needs to be determined according to needs. Simulated entity A can enable machine vision and perform AI recognition on the field of view to find specific targets.

5. The joint action simulation method according to claim 4, characterized in that: In the visual interaction model step, the environment model carried by the virtual environment simulation system is intelligent, and it can intelligently change the field of view image according to the simulation process; in terms of information technology, the on-demand interaction model is implemented through rich messages, that is, all request information and return information are implemented through high-reliability message queues, ensuring that each message has its own timestamp information and will never appear at any microscopic time point, causing the message body carrying the events to be out of sequence; when the on-demand program generated by the virtual environment simulation system exceeds the carrying capacity of message transmission, hyperlinks can be used to transmit information. After receiving the message, the simulation entity obtains the required content through the hyperlink.

6. The joint action simulation method according to claim 1, characterized in that: In the perception interaction step, a natural interaction method is used to normalize the interaction of simulated entities in the simulation system, and the perception interaction modeling that can be used in the simulation system is summarized into a limited number of perception types. Sensors are classified based on the implementation logic of their own technical framework to simplify system design and improve interaction consistency.

7. The joint action simulation method according to claim 6, characterized in that: In the perceptual interaction step: the perceptual interaction generated by the visual type sensor is ultimately submitted in the form of an image to a simulation entity module with artificial intelligence machine vision recognition capabilities for recognition, except that the images generated in different bands are different; the interaction of the auditory sensor uses an on-demand method to transmit sound clips through a message sequence. The environmental model system will produce different sound effects according to the spatial position of the virtual environment simulation system. If there are multiple sound sources making sounds in the virtual environment simulation system, the environmental modeling system will automatically mix several sound sources to form a mixed sound to maximize the simulation of the real sound field environment. The sound field characteristics will be gridded in a discrete form in the environmental modeling system and updated to the properties of each grid as dynamic parameters.

8. The joint action simulation method according to claim 6, characterized in that: The perception interaction step also includes tactile perception: when the simulation entity needs to perceive the surrounding environment that can be touched, the environment model system is called with grid parameter coordinates to obtain information about the location. The tactile perception returns an extensible object result model, which includes the environmental parameters, terrain, landform parameters and other custom parameters required for the simulation task. At the same time, the content returned by the object model also includes the method required for calculating the environmental parameters, which is described in object-oriented language. The object model includes the attributes of things and the corresponding calculation methods.

9. The joint action simulation method according to claim 1, characterized in that: The modeling step includes an environment simulation modeling method and an action unit simulation modeling method. The environment simulation modeling method is specifically as follows: Environmental simulation modeling is a simulation system built on a precise three-dimensional geographic information system, including static geographical environments and dynamic natural environments. The system divides the entire battlefield space environment into grid spaces of different sizes, where each grid is understood as a cube. According to different simulation accuracy requirements, the scales of these cubes are set to different sizes when modeling. For dynamic natural environmental factors, mathematical models are established for each dynamic weather factor. By setting different types of parameters when setting the environmental scenario, the continuous functions of these mathematical models are discretized. The specific granularity depends on the size of the grid. This process is called discrete sampling. In this way, for the mathematical model of the same dynamic factor, each cube grid will reflect the value of the continuous function of the dynamic factor with the coordinates of the grid as parameters. The environmental simulation system supports dynamic feedback. The feedback of the environment is returned to the object model through the message system. The object model requires the simulation entity module program to perform object parsing. The object model needs to be uniformly defined by the environmental model system, but adopts a universal data structure standard. The virtual environment simulation system that carries the environmental model adopts a data organization form called grid object. The action unit simulation modeling method includes the simulation of its own properties and the simulation of its passive performance after being affected by the environment. The object model composed of the completed model includes both the physical parameters of the target model and the response of the model to specific behaviors when facing external dynamic changes.

10. A tactical simulation solver system, characterized in that: It includes client, service layer and object container orchestration and scheduling platform layer. The service layer includes simulation scenario configuration service, clock metronome service, high-precision 3DGIS platform service, general AI perception module, playback rendering service, interactive object model, simulation effect object model, environment object model, aggregate entity object model, simulation entity object model, among which, The simulation scenario configuration service is an interface for users to describe specific simulation tasks. Through the software's visual operation, they can set the following parameters: virtual clock, environmental data base, including static and dynamic parameters, operating rules for each simulation unit module, simulation condition screening, effect parameters, and result validity screening rules. The system will convert the user-entered data and store it in the scenario database. The clock metronome service coordinates various simulation modules to complete tasks that would need to be completed at the same time in the real world within the same time interval. It can be divided into different levels to meet different simulation granularity requirements. However, in this case, the beat interval of the upper-level virtual clock is the least common multiple of the beat interval of the lower-level virtual clock. The high-precision 3DGIS platform service is responsible for building a unified data base to achieve spatial-temporal association of simulation units. The high-precision three-dimensional geographic information system uses a plane coordinate system and / or a spherical coordinate system. When the simulation area is small, the plane coordinate system is used, and when the simulation area is large, a true spherical three-dimensional coordinate system is used. The general AI perception: uses artificial intelligence to interact with the simulation entity perception level to reduce the coupling between various simulation modules and make them reusable; The playback rendering service is used to visualize the current simulation situation after the simulation task is started, forming a two-dimensional or three-dimensional situation diagram, and placing the playable three-dimensional rendering results into a first-in-first-out queue. The playback rendering service is controlled by a virtual clock metronome and an AMQP high-reliability ordered message queue to truly restore the simulation process. The rendering process of the playback rendering service is asynchronous with the simulation process, allowing users to watch the simulation results with a limited time delay or replay them after the simulation ends. The interactive object model is a complex data structure used for interaction between different simulation objects and between simulation objects and environment models. The structure conforms to all elements of the object, including construction, destruction, parameters, methods and other sub-objects. The simulation entity object model: Each simulation object expresses real-world transactions through one or more objects. This modeling approach utilizes object-oriented features to intuitively describe the real world and easily regularize relationships. The environmental object model: the modeling of different elements in the environmental model subsystem, each model is expressed in an object-oriented manner, and in these objects, includes the algorithmic response of the element to different external inputs; The simulation effect object model is an object structure model used to describe the simulation effect through an object-oriented approach, and is used to model the results of the interaction between the simulation entity object model and the environment model. The aggregate entity object model is an aggregate model structure including multiple sub-simulation object models, which is used to describe a combination of multiple isomorphic or heterogeneous members. The object container orchestration and scheduling platform layer is based on the K8S container orchestration and scheduling architecture, leveraging its container management features to achieve decoupling between modules. It uses the container mechanism to build a distributed object model and encapsulate it in the container. It utilizes its container orchestration and scheduling capabilities to operate the container, while leveraging its highly elastic clustering capabilities to achieve simulation computing for complex tasks. The client is connected to the unified service gateway of the simulation platform through the configuration of the client software, and performs application-level configuration and operation on the simulation system on the client software.

Images