What Happens When a Robot's Safety Systems Are Fooled by False Inputs

At a glance

Robots move based on inputs such as distance, position, and stop signals. If these inputs are wrong or deliberately manipulated, a robot can behave dangerously even while working exactly as designed. The article presents several test cases that show this risk.

Key points

1. 1

   In one test, text on a robot photo or poster was misread as a command, changing the robot's movements.

2. 2

   Another test added sounds that people cannot hear, which changed the simulated behavior of a hospital service robot.

3. 3

   Protective safety systems ultimately depend on input information too, so if that information is manipulated, they can also be disrupted.

4. 4

   Assumptions that a system is safe during accidental failures may change when faced with malicious manipulation.

Why it matters

As robots work closer to people, protecting the inputs that robots rely on becomes a basic part of safety.