T/JSIA 0003-2024
Specification for Decision-Making System Based on Multi-Agent Reinforcement Learning (English Version)

Standard No.
T/JSIA 0003-2024
Language
Chinese, Available in English version
Release Date
2024
Published By
Group Standards of the People's Republic of China
Latest
T/JSIA 0003-2024
 

Scope
4 General Requirements 4.1 System Composition The system consists of a data storage module, an information configuration module, an interaction display module, and a strategy generation module. 4.1.1 Data Storage Module Database systems or file systems are used to achieve persistent storage of sensing data and algorithm models. 4.1.1.1 Basic Technical Requirements for the Data Storage Module a) Database systems or file systems should be used to implement persistent storage of data; b) Appropriate data structures and storage methods should be selected to support efficient data querying and retrieval. 4.1.1.2 Functional Requirements for the Data Storage Module a) Provide write and read interfaces, allowing decision-making systems to store critical data in a database or file, and retrieve data from it; the storage system’s central storage should support dynamic and continuous expansion of storage capacity; b) Support batch storage and querying of data to facilitate training and analysis of large-scale data, and possess redundancy backup functions. 4.1.1.3 Performance Requirements for the Data Storage Module a) Concurrent storage performance: The data storage should meet the requirements for concurrent writing; b) The system should not exhibit abnormal phenomena due to response delays from the storage system; database query efficiency should be high enough to return results within a short period of time; c) Storage devices should optimize cache design, adopting a strategy prioritizing record performance to ensure real-time data recording. Data persistence should be supported to ensure that data is not lost when the system is shut down. 4.1.2 Information Configuration Module Scenario information for applications of reinforcement learning algorithms can be custom configured: such as the number of agents and basic environment configuration information. Realize user interface technologies, like graphical user interfaces (GUI) or command-line interfaces (CLI), using user interface libraries or frameworks to create a friendly configuration interface. 4.1.2.1 Basic Technical Requirements for the Information Configuration Module a) Design interfaces that allow users to configure parameters, algorithms, models, etc., of the decision-making system; the system's information configuration function should support dynamic settings of environment basic property information; b) The internal modules of the system should have high cohesion and low coupling characteristics, ensuring data content security and stability. 4.1.2.2 Functional Requirements for the Information Configuration Module Users interact to configure system environment information, initialize the combat environment, provide diverse and reliable operational environments for reinforcement learning algorithms; user interface response times should be quick, with no noticeable delays when users operate on the interface. 4.1.2.3 Performance Requirements for the Information Configuration Module Scalability: The hardware environment should support dynamic expansion of interaction environment scale; the system should not experience visual interface stuttering or other anomalies due to complex environments. 4.1.3 Interaction Display Module Environment visualization: Show current environmental states, which may be images, maps, or other visualization interfaces, allowing users to observe the state of the system. This helps users understand the situation in the environment and monitor the execution of decision-making systems. Strategy display: In the environment visualization, show the strategies currently being executed by various parties (agents, entities, etc.), allowing users to observe their decision behaviors in real-time. This can be achieved through arrows, movement trajectories, etc. Real-time data updates: The interaction display module must be capable of acquiring real-time data, including environmental states, actions of individual agents, scores, etc., which will be used to update the environment visualization and strategy display. Statistical information display: Display statistical information on various indicators in the environment, such as scores, win rates, execution counts, etc. This helps users better understand system performance. Interactive interface: Provide ways for users to interact with the display module, such as buttons, sliders, etc., allowing users to change environmental parameters or adjust the display method. Data recording and replay: Record displayed data so that users can replay and analyze previous decision-making processes at any time. This is very helpful for improving strategies and analyzing failure cases. Configurability: Allow users to customize the layout, content, and style of the display module according to their needs to adapt to different application scenarios. Error handling and notifications: Provide appropriate error handling mechanisms and notification messages so that users can be promptly informed when issues arise. Multi-mode support: Support various display modes such as real-time mode, replay mode, comparison mode, etc., to meet the needs of different users. Scalability: Considering potential future requirements, the display module should be easy to expand and upgrade.

T/JSIA 0003-2024 history

  • 2024 T/JSIA 0003-2024 Specification for Decision-Making System Based on Multi-Agent Reinforcement Learning
  • 2023 T/JSIA 0003-2023 Alternative Testing Specification for Domestic Relational Databases
  • 2021 T/JSIA 0003-2021 Jiangsu Provincial Blockchain Industry Talent Training Base Evaluation Specifications (Trial)
  • 2020 T/JSIA 0003-2020 Intelligent Decision System Standard Based on Reinforcement Learning

Standard and Specification

DIN EN 301465:2001 Private Integrated Services Network (PISN) - Circuit emulation specifications - Emulation of Basic Access by ATM Networks (Endorsement of the English version EN 301465 V 1.1.1 (2000-06) as German standard) DL/T 1024-2006 Technical specification for hydroelectric power plant simulator GJB 8069.1-2013 General requirements for weapon equipment simulation training software.Part 1: General GJB 8205-2014 General requirement for equipment model of EW warfare simulation GJB 7099.3-2012 General requirements for warfare simulation modeling.Part 3: Simulation program modeling GJB 7870-2012 General requirements of space environment data for modeling and simulation GJB 7860-2012 General requirements for simulation management model GJB 7872-2012 General requirements for simulation model service GJB 7873-2012 General requirements for maintenance and upgrade of simulation model GB/T 32297-2015 Aerospace Control System Simulation Requirements GJB 9422-2018 General requirements for force entity behavior simulation models GJB 9428-2018 General requirements for behavioral simulation models GJB 9226-2017 General requirements for verification, verification and validation of military modeling and simulation models T/JSIA 0003-2023 Alternative Testing Specification for Domestic Relational Databases T/CCSA 335-2021 The capability classification requirements for database application migration services T/SDIE 18-2023 Domestic database standard specifications YD/T 4707-2024 Cyberspace Security Simulation Smart Car Security Simulation Platform Access Requirements T/JSIA 0003-2020 Intelligent Decision System Standard Based on Reinforcement Learning GJB 6935-2009 Military simulation terminology



Copyright ©2025 All Rights Reserved