Pain is painful, but it helps. It tells us that there is a problem and that we must react. Researchers at the Technical University of Singapore devised a system that would give robots a similar feel. The goal is for the robot to recognize a “painful” situation and to be able to put itself out of danger.
Scientists have created sensors based on “memtransistors” – a set of transistors that simulate a memristor. The latter is a theoretical passive electronic component whose resistance changes as a function of the applied current and which makes it possible to store information. Its advantage is that it can simulate the behavior of neurons and synapses. In the present study, these memtransistors are used to learn to recognize a physical event that breaks or damages robotic functionality.
Scientists have also connected near this “artificial skin” artificial intelligence devices to interpret these pain signals and trigger a reaction in real time. Example: a knife rests on the hand of a robot who recognizes the danger and withdraws it. Compared to a centralized architecture, these “mini-brains” placed very close to the sensors have the advantage of being able to provoke a faster reaction. In addition, they avoid a lot of wiring.
And that’s not all. Researchers have also developed an “ionic gel” which allows, to a certain extent, to automatically repair certain damage which would have appeared at the electronic level. A bit like when a biological body heals itself after a small injury.
Source : Singapore Technical University