Expert Systems: AI-Assisted Battlefield Decision-Making

Authors: Zhang Yuan, You Jiatong

Source: China Military Network

In modern battlefields, weaponry is becoming increasingly advanced and complex. When encountering problems during use, soldiers often turn to the experts who design or manufacture these weapons. But what if the experts are not available? An applied artificial intelligence system—”Expert System”—has emerged to address this issue.

When it comes to artificial intelligence, most people might think of AlphaGo, which defeated human professional Go players. However, expert systems are different from the AI systems people typically envision. They do not solve academic problems that can be abstracted into mathematical models, like playing Go, but rather address specific practical difficulties in narrow fields. For instance, they can be used for monitoring helicopter engine rotors or assigning targets for air defense weapons… These problems often require human experts to solve based on years of practical experience. Expert systems simulate human expert thinking, enabling machines to solve complex real-world problems that only experts can typically address in a short time.

Conceptual diagram of expert systems. Created by Wang Mengyuan

Expert systems typically possess a vast knowledge base. There are generally two ways to acquire knowledge: one is for human experts to teach the knowledge and skills they possess to machines using tools like knowledge editors, similar to how a master teaches an apprentice; the second is through the system’s own autonomous learning capabilities, summarizing a large amount of feedback information during operation to generate new knowledge. The knowledge base is significantly different from traditional databases; the knowledge within is not passive or static but active and dynamic. Knowledge base management systems can modify and maintain the knowledge at any time, making the knowledge base increasingly rich and complete. The level of the knowledge base directly determines the capability of an expert system. In military fields, represented by combat command, knowledge often manifests as a large amount of experience and intuition that can only be understood and not easily articulated, requiring long-term effective accumulation and transformation during training and wartime.

Unlike the traditional problem-solving process of “data structure + algorithm,” expert systems solve problems through “knowledge + reasoning.” The inference engine is the brain of the expert system, responsible for mimicking the thought process of experts and executing and controlling the problem-solving process. The inference engine utilizes knowledge for reasoning, but its performance is minimally affected by the content of the knowledge base. For example, a commander may be able to recite training regulations verbatim, but that does not mean they can effectively organize training. Expert systems possess rich knowledge, and the thinking methods are diverse, including precise reasoning, uncertain reasoning, incomplete reasoning, and exploratory reasoning. Of course, traditional systems produce precise answers every time based on algorithms. In contrast, expert systems think like humans and may inevitably produce incorrect answers. However, the system can learn from its mistakes and continuously improve.

Expert systems are generally interactive “human-machine” systems. Through the human-machine interface, users can inquire through dialogue, and the machine uses the knowledge it possesses to reason based on the facts provided by the user, ultimately providing answers to the questions. It is worth mentioning that expert systems can not only provide methods for solving problems but also answer the user’s questions of “why,” explaining how they arrived at their conclusions. This is a relatively transparent communication process that not only builds trust between humans and machines but also helps users quickly identify and correct system errors.

Today, expert systems have become a relatively mature branch of artificial intelligence. Upholding the principle of practicality, they have begun to be applied in the military field, such as in fault diagnosis and maintenance guidance for various weapon systems, military transport scheduling management, combat mission planning, and intelligence analysis. Their application prospects and development space are vast, and they are likely to become a new tool to assist in winning future battlefields.