Exercises Modernization Doctrine

A Glimpse at the Transformation of Military Training Organization Methods

管窥军事训练组训方式之变

PLA Daily (解放军报) 11 June 2026

Summary

A PLA military training journal article outlines five emerging training organization methods the force is developing to align with informatized and intelligentized warfare: geographically distributed training, scenario-immersive training, unmanned autonomous training, human-machine collaborative training, and virtual-physical interactive (LVC) training. The article documents a doctrinal problem the PLA is actively working through—how to restructure training infrastructure and methodology when the combat system increasingly depends on autonomous platforms, AI-driven decision support, and distributed simulation rather than co-located live forces. The explicit discussion of transitioning from 'human in the loop' to 'human on the loop' and 'human out of the loop' confirms that delegating authority to unmanned systems is not merely a capability aspiration but a training design problem the institution is now formalizing.

Translation

Military training organization methods (军事训练组训方式) are the methods and forms adopted to organize and implement military training. At present, as high-technology accelerates the evolution of the form of warfare (战争形态) toward informatized and intelligentized warfare (信息化智能化), it is also quietly driving military training organization methods to transform in the directions of "network+," "digital+," and "intelligent+," broadening the methodological pathways for improving the quality and efficiency of combat capability generation in the force.

Geographically distributed training (异地分布式训练) refers to a training organization method that employs wireless communications, the Internet of Things, virtual simulation, and other technologies to build a distributed simulation system with full-domain coverage, end-to-end connectivity, and dynamic interaction—transcending physical space, breaking venue limitations, and freeing training from environmental dependencies—in order to organize multi-force real-time synchronized training across different locations, truly realizing large-scale exercises and training that span multiple echelons within a single virtually synthesized operational scenario. For example, foreign militaries have focused on developing plug-and-play auxiliary devices that allow trainees stationed at different locations to connect to a common synthetic training environment using only a single network plug-in. To conduct geographically distributed training, one should rely on command information systems to build a live-equipment operational environment, construct exercise control and simulated engagement environments, open up remote information links, and integrate and connect operational systems with training systems—creating an integrated exercise and training environment support structure that is distributed across multiple remote locations, synchronized across time and space, and enables real-time data flow—so that participating forces can engage in operational simulation training unconstrained by time, space, or environment, making full use of dynamic distributed interaction functions to increase training difficulty and intensity, and achieving the goals of validating operational theories and plans, predicting the effects of operational actions, assessing weapons system performance, and inspiring innovative operational thinking.

As the level of simulation technology continues to rise, training methods that use military games, live-environment equipment, and other means to simulate the real battlefield have emerged accordingly, and can greatly improve training quality and effectiveness while reducing training costs and training injuries. By employing virtual reality, augmented reality, mixed reality, and other technologies together with wearable devices, a virtual battlefield environment that closely approximates reality and is fraught with danger can be constructed, enabling trainees to perceive images, sounds, smells, and other sensations close to those of a real battlefield, achieving an experience of immersion in actual combat scenarios that transcends real-world sensory perception, and realizing comprehensive simulation from decision-making simulation to fire strikes and from individual soldier actions to the overall linkage of the combat system. For example, a digital individual-soldier immersive training system (数字单兵沉浸式训练系统) can provide simulated virtual battlefield environments including mountain, jungle, desert, and other scenarios; soldiers participating in training need only carry various sensor devices and select different battlefield environments and mission plans to experience the effects of live-combat training within the system. To conduct scenario-immersive training (场景沉浸式训练), one should design key actions such as manned-unmanned collaboration around operational tasks, immersing trainees in realistic scenarios to conduct multi-perspective, unrestricted, and first-person reconnaissance of typical battlefield targets and key information, understand the performance of the adversary's main combat weapons and equipment employment methods, and train repeatedly while boldly accepting errors, achieving the goals of optimizing tactics and achieving precise coordination.

The deep application of big data, artificial intelligence, unmanned technology, and other fields in the equipment domain has endowed unmanned equipment, command information systems, and other systems with autonomous learning capabilities and intelligent cognitive functions. Unmanned intelligent equipment such as unmanned aerial vehicles, unmanned tanks, unmanned vessels, and nano-robots, drawing on powerful storage and computing capabilities, build networks for deep learning and repeated training by learning from human knowledge and experience, accumulating massive volumes of exercise and training data, and continuously optimizing algorithm models, enabling deep training of weapons and equipment systems and autonomous enhancement of combat capability. For example, by using artificial intelligence military analysis systems to collect and deeply analyze large volumes of combat video from the battlefield front and "feed" unmanned intelligent equipment models, such systems can rapidly absorb and digest changes in the battlefield situation, precisely identify targets, and assess weapons system effectiveness. To conduct unmanned autonomous training (无人自主式训练), one should innovate operational command groupings such as "human-machine integration (人机一体)," "authorized division of labor (授权分工)," and "unmanned autonomy (无人自主)," with emphasis on strengthening training in operational planning capabilities (运筹能力) such as situational analysis, requirements forecasting, and plan deduction; support capability training such as autonomous identification of equipment faults, precise prediction of materiel requirements, and rapid drafting of maintenance plans; and defensive capability training such as autonomous detection, warning, and attribution of cyber-electronic attacks.

On the future informatized and intelligentized battlefield, manned and unmanned combat platforms will typically operate in coordination in key areas and phases, and the level of human ability to control unmanned intelligent weapons will directly affect the realization of weapons and equipment effectiveness. Only by strengthening coordinated training between manned and unmanned combat platforms, mastering the operational characteristics of unmanned combat platforms and the laws governing human-machine coordinated training, and achieving dynamic task allocation and sharing between humans and machines as well as role exchange and fluidity, can seamless integration of human-machine actions and complementary synergy of human-machine advantages be achieved. To conduct human-machine collaborative training (人机协同式训练), on one hand, one must continuously improve human intelligentized literacy (智能化素养) through deep interaction with intelligent machines, improve the quality of human-machine cognition, understanding, and interaction, and enhance the human capacity to master intelligentized combat systems. On the other hand, one must explore new training models with the "machine" as the primary subject, transfer human experience, enhance machine intelligence, and accelerate the transition from "human in the loop (人在回路中)" to "human on the loop (人在回路上)" and "human out of the loop (人在回路外)."

Based on technologies such as big data, cloud computing, digital twins, and virtual-physical interactive perception, mapping geographically distributed, differently configured physical equipment, hardware-in-the-loop simulation equipment, and digital simulation models into a virtual battlefield space that is consistent in time and space can achieve physical interconnection and interoperability, real-time interactive action, and coupled transmission of effects, forming a system-of-systems training environment (体系训练环境) that is "logically integrated and virtually-physically combined (逻辑一体、虚实结合)," enabling deduction and training exercises that are virtually-physically combined and highly interactive under unified operational condition drivers. For example, an integrated training environment based on Live, Virtual, and Constructive (LVC) simulation can achieve the fusion and integration of simulated training means with live-combat means, efficiently completing training in weapons and sensor operation, tactical decision-making, special situation handling, and other areas, greatly improving training quality and effectiveness. To conduct virtual-physical interactive training (虚实交互式训练), on one hand, under conditions of sufficient data and models of operational adversaries, one should conduct large volumes of interactive training against different operational adversaries through virtual battlefield environments, researching and practicing strategies for victory, and using the virtual to promote the real (以虚促实). On the other hand, one should train artificial intelligence training systems by mapping physical data, causing them to continuously iterate, evolve, and upgrade so as to become more intelligent and more efficient, using the real to promote the virtual (以实促虚).

Original Chinese

军事训练组训方式，是组织实施军事训练所采取的方法和形式。当前，高新技术加速推动战争形态向信息化智能化演进的同时，也悄无声息地牵引军事训练组训方式向“网络+”“数字+”“智能+”方向转变，为部队战斗力生成提质增效拓宽了方法路径。异地分布式训练指运用无线通信、物联网、虚拟仿真等技术，搭建覆盖全域、贯通末端、动态交互的分布式仿真系统，跨越物理空间、打破场地限制、摆脱环境依赖，组织多部队异地实时同步训练，真正实现在同一个虚拟合成的作战场景中，进行跨越多级的大规模演训的组训方式。比如，外军聚焦开发即插即用型辅助设备，使驻扎在不同地点的受训对象只需使用一个网络插件，即可连接到共同的合成化训练环境之中。开展异地分布式训练，应依托指挥信息系统构建实装作业环境、构设导调和仿真交战环境、打通远程信息链路，集成链接作战系统与训练系统，打造异地多点分布、时空远程同步、数据实时流转的一体化演训环境支撑，使参训部队不受时空环境限制，参与到作战模拟训练之中，充分利用动态分布交互功能增加训练难度和强度，达成检验作战理论和计划、预测作战行动效果、评估武器系统性能、启发创新作战思想等目的。随着仿真模拟技术水平的不断提高，通过军事游戏、实景器材等模拟展现真实战场的训练方式应运而生，可在节约训练成本、减少训练损伤的同时，极大提升训练质效。运用虚拟现实、增强现实、混合现实等技术和可穿戴设备，构建逼近真实、险象环生的虚拟战场环境，使受训者能够感触到接近真实战场的画面、声音、味道等，达到沉浸实战场景、超越现实感官的体验，实现从决策模拟到火力打击、从单兵行动到作战系统整体联动的全面仿真模拟。比如，数字单兵沉浸式训练系统，能够提供包含山地、丛林、沙漠等场景在内的仿真虚拟战场环境，参加训练的士兵只需携带各种传感设备，选择不同的战场环境和任务方案，就能在系统中体验到实战训练的效果。开展场景沉浸式训练，应围绕作战任务设计有人无人协同等关键行动，让受训者沉浸在逼真场景中，多视角、无限制、代入式勘察典型战场目标和关键信息，了解对手主战武器性能和装备运用方法，反复训练、大胆试错，达成优化战法和精确协同的目的。大数据、人工智能、无人技术等在装备领域的深度运用，使无人化装备、指挥信息系统等具备自主学习能力和智能认知功能。无人机、无人坦克、无人舰艇、纳米机器人等无人智能装备，基于强大存储与运算能力，通过学习人类知识经验、积累海量演训数据、不断优化算法模型，构建起用于深度学习、反复训练的网络，能够实现武器装备系统深度训练，自主提升战斗力。比如，利用人工智能军事分析系统，大量收集并深度分析来自战场前线的作战视频，“喂养”无人智能化装备模型，可使其快速消化吸收战场变化情况，精确识别目标和评估武器系统有效性。开展无人自主式训练，应创新“人机一体”“授权分工”“无人自主”等作战指挥编组，重点强化态势分析、需求预计、方案推演等运筹能力训练，自主判别装备故障、精确预测器材需求、快速拟制维修计划等保障能力训练，以及对网电攻击自主探测、告警与溯源等防御能力训练。未来信息化智能化战场上，在关键地域和阶段通常是有人与无人作战平台协同行动，人类操控无人智能武器的能力水平，将直接影响武器装备效能的发挥。只有加强有人与无人作战平台的协同训练，掌握无人作战平台行动特点和人机协同训练规律，实现人与机器的任务动态分配与共享、角色互换与流动，才能达成人机行动无缝对接、人机优势协同互补。开展人机协同式训练，一方面，要通过与智能机器的深度交互不断提升人的智能化素养，改善人机认知、理解、交互质量，提高人驾驭智能化作战系统的能力。另一方面，要探索以“机”为主体对象的新型训练模式，迁移人类经验，提升机器智能，加快实现由“人在回路中”向“人在回路上”“人在回路外”的转变。基于大数据、云计算、数字孪生、虚实交互感知等技术，将异地分布、形态不同的实体装备、半实物模拟装备和数字仿真模型映射到时空一致的虚拟战场空间，可实现物理上互联互通、行动上实时交互、效果上耦合传递，形成“逻辑一体、虚实结合”的体系训练环境，在统一作业条件驱动下开展虚实结合、高效互动的推演训练。比如，基于“真实、虚拟、构造”（LVC）的一体化训练环境，能够实现模拟训练手段与实战手段的融合一体，可高效完成武器与传感器操作、战术决策、特情处置等训练，极大提升训练质效。开展虚实交互式训练，一方面，要在作战对象数据和模型充足条件下，通过虚拟战场环境与不同作战对手进行大量交互训练，研练制胜之策，实现以虚促实。另一方面，要通过映射实体数据，训练人工智能训练系统，使其不断迭代演化升级，变得更智慧更高效，实现以实促虚。