China’s latest Five-Year Plan makes a short-term objective clear: cbecome the first world power. This covers numerous areas such as energy (both renewable and nuclear), the technological with AI, robotics and the development of its own chips, education thanks to new technologies and the military. Curiously, they are all intertwined and there is something that has been playing for a while: futuristic weaponry.
Like the rest of the powers, China does not hesitate to show its military potential, but in recent months we are seeing that the discourse is focused on capabilities that, until not so long ago, seemed more typical of the field of science fiction. The latest is a technology for a swarm of drones to be able to orchestrate autonomously on the battlefield with a single objective.
Hunt and destroy enemies until not a single one remains.
HG-STR. Dubbed ‘Heterogeneus Graph Spation-Temporal Reasoning’, or HG-STR, we are talking about an algorithm that would be the brain of a fleet of fixed-wing drones that would not need humans to operate. Currently, most operations involving drones still require a human at the controls (sometimes those controls are such everyday objects like a Steam Deck either the xbox controller). However, the HG-STR would represent a paradigm shift.
According to an undisclosed source SCMPthis technology opens the doors to a future in which swarms of drones could be sent into a high-risk hostile environment in which there is no contact with human operators, but there is a clear order in the programming: eliminate all enemies.
Rule change. Currently, hybrid or “traditional” models operate with a single database that unites information on allies, enemies and the terrain in which they operate. In different environments when drones operate autonomously, this creates confusion and that is why a human is needed to take the final command. With this development, things change.
The algorithm has different ‘sections’ or mailboxes to which it sends the information it has to process. Instead of operating with a single database, it makes decisions based on whether an ‘object’ is friend, foe, or a search area. In the case of being an ally, it does nothing; If it is a search area, it strives to find the enemy; If it’s an enemy, shoot.
According to one of the authors of the study published in China’s leading peer-reviewed aviation journal, “this allows the swarm to instantly understand who to help and who to hunt. This adaptability is important because rules-based systems fail when the enemy does not follow as expected, while HG-STR is able to adapt.”
order in chaos. Something key here is speed. The researcher points out that, when a drone is in combat, it is too slow in making decisions. “In the heat of battle, they take seconds to decide, a time in which an unmanned aircraft can fly almost 600 meters blindly, representing a fatal delay in electromagnetic warfare.” HG-STR, however, makes decisions in just 6.6 milliseconds. Practically in real time.
It is this chaos where the team of researchers wanted to focus thanks to an interesting solution: providing each drone with a “memory”. Although there is a central algorithm, if one of the drones loses contact with its companions, it ‘pulls’ the memory to remember where its allies were before losing contact and where they last saw enemies.
Once those priorities are sorted, the drone searches for its objective and another decision comes into play: do I attack or continue searching? Once this is done, you choose a specific target and finally decide how much ammo you need to take it down. Instead of having one set of general instructions, the drone software divides problems into layers, avoiding clutter by having to process everything at once.
“Kill them all.” The study notes that HG-STR is the first known algorithm capable of achieving a 100% kill rate while operating autonomously and fast enough to react in real time to the rapidly changing conditions of a modern warfield.
All this is scary, but the most terrifying thing is that, according to the experiments, the researchers carried out different simulations in which they tested this autonomous system. In complicated scenarios where they limited communication systems, they claim the algorithm achieved a 100% kill rate on enemy targets, including those hidden in plain sight.
They are now focusing on scaling the system, as they have realized that the algorithm can be adapted to other contexts of larger battlefields, more targets and more drones simultaneously without needing to retrain the AI.
Context. As I say, this study does not arrive in a vacuum, but in the context of China’s acceleration towards autonomous drone warfare. A few months ago we already echoed the command of robotic “wolves” who were already doing maneuvers alongside flesh and blood soldiers, but over the last two years we have witnessed other demonstrations in which individual soldiers can control a couple of hundred drones to operate autonomously, as well as other robotic weaponry and even ‘ship’ concepts that seem straight out of ‘Star Wars’.
It is, in short, one more step towards what is already known as war without human intervention in which machines are the ones making the decisions independently. And, far from being a private initiative, this HG-STR has been funded by the National Natural Science Foundation of China, which gives an example of what we said at the beginning of the article: within the Five-Year Plan, everything is connected.
Just imagine if the great powers put all this technology into place to meet other humanitarian goals rather than to find more efficient and effective ways to take each other out.
GIPHY App Key not set. Please check settings