According to foreign media reports, the British Army is intensifying the development of an AI target positioning system called "Asgard". The system relies on advanced communication networks and new algorithms to complete threat target monitoring and locking within minutes, greatly improving the efficiency of remote strikes. At present, the project has received funding support of over 1 billion pounds (approximately 1.345 billion US dollars) from the UK Department of Defense. Currently, AI target positioning systems have become a key area of military technology research and development in many countries due to their high efficiency and accuracy. The AI target positioning system from information extraction to intelligent decision-making is an intelligent system that utilizes computer vision, sensor fusion, and AI algorithms to achieve automatic recognition, spatial coordinate calculation, and attribute classification of specific targets. The core process is that the system obtains environmental perception data through sensors (cameras, radars, etc.), and then uses models for data analysis and processing to identify preset category targets and calculate the precise location information of the output targets. As early as the Cold War, there were studies on target detection technology. Under the technological conditions at that time, this type of technology mainly relied on hard coded rules and simple pattern matching for basic detection and recognition, with very limited information processing capabilities. For example, early satellite reconnaissance and identification systems could only recognize large, high contrast targets such as missile launch pits, and were often helpless when facing complex environments or disguised targets. From the end of the 20th century to the beginning of this century, with the development of machine learning technology, AI target positioning systems have made significant breakthroughs. During this period, such systems began to have the ability to learn target feature patterns from data. The relevant systems deployed by the US military during the Kosovo War were able to combine visible light and infrared images to perform probabilistic recognition based on pre input target features, which to some extent improved the processing speed and intelligence output efficiency of target recognition. However, due to the limitations of the algorithms and data scale at the time, the recognition accuracy and generalization ability of these systems (the adaptability of machine learning algorithms to new data) were still limited, and largely relied on manual assisted recognition. The turning point that truly drives revolutionary changes in AI target positioning systems is the breakthrough in deep learning technology at the beginning of the 21st century, represented by the rapid development of its classic algorithm - convolutional neural network technology. This algorithm can extract image features at multiple levels, allowing the computational model to recognize targets that have been disguised or partially occluded within a certain range. At the same time, the development of computing hardware and the emergence of large annotated datasets (important resources for machine learning training, helping machines improve model performance and generalization ability through massive data learning) have provided a foundation for training complex deep learning models. At this stage, AI target positioning systems are gradually transitioning from simple information extraction to intelligent decision-making. Reconstructing the traditional kill chain At present, the research of AI target positioning system focuses on three major directions - multimodal data fusion, edge computing and system autonomy improvement. The AI target positioning system is no longer limited to the analysis of a single data source, but integrates multiple sources of information such as visible light, infrared, synthetic aperture radar, electronic signals, open-source intelligence, and acoustic data to construct a panoramic target situation map, providing support for combat decision-making. For example, the AI target positioning system previously tested by NATO can correlate and analyze drone videos, radio intercept signals, and social media information to achieve precise positioning of high-value targets. The so-called edge computing refers to processing information close to the source of data generation. The development of this distributed computing framework makes information processing sink from the cloud to the tactical edge (such as unmanned aerial vehicles, combat vehicle platforms), realizing near real-time links from sensors to shooters. If the drone collects data in real-time during task execution, transmits the data to the cloud for processing, and then returns, there may be delays. Edge computing allows UAVs to use airborne equipment to directly process data and make real-time responses. This not only improves response speed, but also reduces bandwidth requirements and latency for data transmission, making the system more efficient and reliable. At the same time, research teams from multiple countries are working to enhance the autonomous capabilities of equipment systems. For example, the US Air Force Research Laboratory is advancing the "Golden Horde" autonomous collaborative guided ammunition project. This project utilizes networking technology to enable unmanned aerial vehicles to collaborate autonomously without relying on human intervention, thereby improving combat efficiency and flexibility. The breakthrough application of new technologies is causing AI target positioning systems to reconstruct traditional kill chains. Through real-time processing of multi-source heterogeneous data, rapid response of edge computing, and deep embedding of independent decisions, AI target positioning system can quickly locate opponent communication nodes, radar emitters or key network nodes, providing important support for electronic warfare and network warfare. The combat process from discovery to decision-making and then to strike has been compressed to the minute level, improving efficiency by dozens of times compared to traditional modes. The UK's "Asgard" AI target positioning system is a concrete manifestation of this trend. It should be noted that current AI target localization systems heavily rely on complex deep learning techniques to overcome technological challenges and cognitive breakthroughs. Due to the complex architecture of deep learning algorithms, their decision-making process is difficult to understand and track, resulting in a lack of transparency and interpretability in the conclusions drawn by the system. This situation is prone to the "black box" effect, where the decision-making process of AI is like a mysterious black box, and people cannot understand the logic and basis behind it. For example, commanders have no way of knowing how the system identifies a target as an important military facility rather than a civilian building. Data dependency is another key weakness of AI target localization systems. The system heavily relies on multi-source heterogeneous training data, and its recognition accuracy needs to be based on a large amount of high-quality and accurately labeled specific scene data. However, the scarcity of actual combat data makes it difficult for training datasets to fully cover complex battlefield environments. In addition, the data annotation process is time-consuming, labor-intensive, and difficult to standardize, and its complexity also puts higher demands on annotators. It is worth noting that data quality defects or labeling deviations may cause system model inaccuracies. The enemy can induce the system to generate false target heat maps through data pollution, algorithm model attacks, communication interference, and other means, leading to command personnel misjudging the battlefield situation. To solve this dilemma, it is necessary to simultaneously promote the development of battlefield data generation technology, algorithm capabilities, and network security protection system, ensuring that AI target positioning systems maintain efficient recognition capabilities in complex data environments and adversarial threats, and become the "shining eyes" of the future battlefield. (New Society)