It turns out that a healthy dose of fear may be as beneficial for robots as it can be for humans. Researchers were able to teach a robot to enact self-preservation strategies while navigating dynamic environments, and it made the bot better at assessing risks and avoiding danger.
To conduct their study, the team used reinforced learning to emulate a fear response in the robot control system as it processes and adjusts to its environment. In simulations, they compared their system to those without a fear response as they made their way through different scenarios with dangerous and nondangerous obstacles. The team found that their fearful robot navigated a smoother and safer path than those that were not afraid.
“Fear is closely related to self-preservation and rapid responses to danger, both of which are critical for adaptive behavior,” said Andrea Usai, a doctoral candidate at the Polytechnic University of Turin in Italy and part of the team that led the study.
