It’s well established that pain is an important alert system in the body. Computers may be able to flash alert signals when something goes wrong, but they don’t have the instinctive, biological compulsion to stop doing that thing that hurts them. This is why robots that work in extreme, dangerous environments might need an analog to pain to help them make decisions in chaotic workspaces, thereby acting to protect themselves or the humans and equipment around them.
Researchers from Leibniz University in Hannover presented an idea for robot pain reactions at the IEEE International Conference on Robotics and Automation (ICRA) last week, after which their work was profiled in the IEEE Spectrum. Their invention is a reflex controller inspired by living creatures, which could be applied on industrial robots.
The same idea has been used before by researchers at Stanford and the University of Rome–La Sapienza to teach industrial robots not to hurt human workers around them. This is collision avoidance, but it isn’t pain. In order to give the robot a “vocabulary” of experiences it could classify as pain, the researchers developed a hierarchy of sensations, from light to strong. The robot arm – in the case of this experiment captured on video, it is a Kuka arm and a BioTac tactile sensor – will react differently to different levels.
According to the paper Johannes Kuehn and Sami Haddadin wrote for ICRA, the robot will retract faster from pain – a.k.a. unusually severe movement or temperature – depending on the severity of the sensation. The stronger the “feeling,” the farther back it moves and the longer it stays there. “Severe pain” includes damage that could completely incapacitate the robot or things around it. To respond to this, the robot goes into a “strictly passive” mode with gravity compensation, keeping it still until an operator can arrive to help. Ideally, this should prevent the robot itself from causing further damage in the case of a catastrophic structural loss.