In one Blog, Facebook says the solution – Rapid Motor Adaptation – is a “breakthrough” in artificial intelligence (AI). It allows robots to walk on sand, mud, hiking trails, tall grass and a pile of dirt “without a single failure” during tests. “He managed to navigate a pile of cement and a pile of pebbles in 80% of the trials, although he never saw unstable or sagging ground, obstructive vegetation or steps during training,” he said. declared the post. “He also maintained his height with a high success rate when moving with a payload of 12 kg, which is 100% of his body weight,” he added.

To do this, the researchers used two branches of AI, called supervised learning and reinforcement learning. Supervised learning depends on the provision of algorithmic data on the situations it will be confronted with, to which it can refer to make decisions. Reinforcement learning is about making the AI ​​interact with the environment and learning from its own decisions. Facebook says the robot has demonstrated “a fundamental skill for all intelligent agents,” which roughly means it was closer to mimicking the way humans adapt to terrain while walking.

The robot is able to determine the amount of friction offered by a surface, the weight it carries, the anomalies of the terrain, etc. and adapt its movement accordingly, and without actually looking at the surface. The second part is important because, if successful, it eliminates the need to put cameras on such robots, thereby reducing the cost of their construction.

According to Facebook, robots that have to walk on a particular terrain are currently specially designed for that type of terrain. For example, robots working in factories have algorithms adapted to this environment. With RMA, Facebook says robots can adapt to the environment “from scratch” by “exploring and interacting with the world.” This is apparently the first such solution in the world.

Researchers at the Norwegian University of Oslo also presented a similar solution in March, intended to allow robots to adapt to their environment. This solution included the use of machine learning (ML) and AI to change the length of the robot’s legs and its body shape to suit the terrain. Essentially, the solution was to adapt the way humans dynamically change their center of gravity while walking across different terrains. Like Facebook, the Oslo researchers also presented robots resembling four-legged dogs, but their solution depended on 3D cameras and force sensors. Facebook hasn’t disclosed what kind of sensors its bots need.