China’s new ‘kill them all’ algorithm lets AI take over drone targeting

Chinese researchers have developed a new artificial intelligence algorithm designed to enhance the targeting capabilities of drone swarms on the battlefield, even under conditions of communication jamming. The AI system, called Heterogeneous Graph Spatio-Temporal Reasoning (HG-STR), enables drones to make rapid and coordinated decisions using incomplete information. This breakthrough allows a swarm of drones to efficiently search large areas and eliminate all enemy targets, potentially achieving a 100 percent kill rate according to the developers from Northwestern Polytechnical University in Xi’an. The HG-STR algorithm operates by mapping each drone, enemy target, and surrounding environment as interconnected nodes within a heterogeneous graph. Each drone node carries data about its position, speed, ammunition, and previous tasks, while enemy targets and unsearched areas are represented as separate nodes with relevant information. This structure allows drones to prioritize threats and opportunities for collaboration, even when communication between units is disrupted by enemy jamming. Additionally, a compressed memory system enables drones to retain past observations and continue reasoning without needing to restart from scratch, enhancing resilience and operational continuity. In practical tests, a swarm of 10 drones equipped with HG-STR successfully patrolled and cleared a 100 km by 100 km area more efficiently and with less flight distance than previous methods. The system’s ability to make decisions within 6.6 milliseconds is particularly significant for fast-paced combat scenarios. This contrasts with existing autonomous drone algorithms, which treat all data uniformly and often require continuous human control, limiting their effectiveness in contested environments. The development of HG-STR marks a potential shift toward fully autonomous drone operations capable of executing complex missions independently, even in communication-denied zones. This advancement could have profound implications for future warfare, enabling drone fleets to carry out coordinated attacks with minimal human intervention and increasing the lethality and survivability of unmanned systems on the battlefield.