Robot fights have become the big trend in the tech sphere lately, and not just to punish children. Beyond the simple spectacle, hand-to-hand combat is a perfect discipline for testing and improving many vital factors in humanoid machines, notably stability, dynamic balance, and movement precision under conditions of physical stress.
Unitree UnifoLM-X2-1.0: AI takes over
So far, the spectacular performances we have witnessed (often viral on social media) were remotely controlled by human operators. The robots did not make decisions themselves... That has now changed. The Chinese company Unitree has recently unveiled the first algorithmic model giving genuine decision-making autonomy to humanoid combat robots. Named UnifoLM-X2-1.0, this model allows the machine to plan its attacks, make tactical decisions, and react in real time to an opponent.
From brawling to domestic assistance
Rest assured, the Asian manufacturer’s ultimate goal is not to limit itself to pure brawling. This is above all an extreme test bench. The real challenge is to use this data to give such a degree of autonomy, fluidity, and responsiveness to all kinds of domestic or industrial robots, so that they can operate in our environment in a completely safe way.
The UnifoLM-X2-1.0 algorithmic model is currently in the testing and internal deployment phase on Unitree's development robots. No public commercial release of these "combat" robots is planned in the immediate future. Impressive...