Chinese startup LimX Dynamics has proven the ability of robots to quickly respond to any emergency and confidently navigate difficult terrain. The design is based on a self-learning model and its ability to respond to external hazards, including moving objects or uneven terrain.
LimX Dynamics uses a separate model to train AI algorithms. In a new test, the company took the P1 robot up Shanghai's Tanglang Mountain. Specialists are testing the system's ability to navigate in an unfamiliar environment with uneven terrain.
Testing requires P1 to be constantly aware of objects around him and then feed that information to a reinforcement learning system. The robot confidently climbs grassy hills, recovering its gait after stumbling. P1 even managed to remain stable after repeated blows to his limbs with a stick.
Reinforcement learning uses rewards or punishments to allow AI to make autonomous decisions. Such algorithms differ from classical artificial intelligence systems in that they are based on autonomous learning.
The startup now aims to apply similar technology to its CL-1 humanoid robot, which can process data from stairs and other objects before successfully traversing them. More about the innovation, see the video.