Robots are becoming more adept at bracing themselves against barriers to prevent falls.
With the aid of AI, bots can behave somewhat intoxicatedly.
So much for
laughing at falling robots. A humanoid TALOS robot's legs are shattered, and
researchers at the University of Lorraine have created a "Damage
Reflex" system (also known as D-Reflex) that causes the robot to balance
awkwardly against a wall. The neural network-based algorithm immediately
identifies a location on the wall that is most likely to offer stability by
using its experience (in this example, 882,000 training simulations). The robot
can reach out roughly as rapidly as a person and doesn't need to know how it
was hurt.
The outcome
is the expected anti-comedy, as noted by IEEE Spectrum. The robot braces itself against the wall like someone who has just strained an ankle instead of
falling to the ground.