A robot in the kitchen cooking you dinner may sound like a great idea. Just make sure you stay out of the way when it is handling the knives.
To find out what would happen if a robot holding a sharp tool accidentally struck a person, German researchers performed a series of stabbing, puncturing, and cutting experiments.
They fitted an articulated robotic arm with various tools (scalpel, kitchen knife, scissors, steak knife, screwdriver) and programmed it to execute different striking maneuvers. They used a block of silicone, a pig’s leg, and at one point a human volunteer’s bare arm as the, uh, test surface.
The researchers—Sami Haddadin, Alin Albu-Schaffer, and Gerd Hirzinger from the Institute of Robotics and Mechatronics, part of the German Aerospace Center (DLR), in Wessling—presented their results today at the IEEE International Conference on Robotics and Automation, in Anchorage, Alaska.
The main goal of the study was to understand the biomechanics of soft-tissue injury caused by a knife-wielding robot. But the researchers also wanted to design and test a collision-detection system that could prevent or at least minimize injury. Apparently the system worked so well that in some cases the researchers were willing to try it on human subjects.
We applaud the man in the video getting his arm struck by the robot in the name of science [UPDATE: That’s Sami Haddadin, the study’s lead author, who was clearly confident in the safety system he and his colleagues devised].
Warning: The video includes scenes that some people may find upsetting.
The researchers acknowledge that there are serious concerns about equipping robots with sharp tools in human environments. But they argue that only by getting more data can roboticists build safer robots.
The experiments involved the DLR Lightweight Robot III, or LWRIII, a 7 degrees-of-freedom robot manipulator with a 1.1-meter reach and moderately flexible joints. The robot, which weighs 14 kilograms, is designed for direct physical interaction and cooperation with humans.
The tools the researchers tested included [right]: (1) scalpel; (2) kitchen knife; (3) scissors; (4) steak knife; (5) screwdriver.
The researchers performed two types of experiments: stabbing and cutting. The actions were done with the different tools at various speeds, with and without the collision-detection system active.
In most cases, the contact resulted in deep cuts and punctures, with potentially lethal consequences. But remarkably, the collision-detection system was able to reduce the depth of the cuts and in a few cases even prevent penetration altogether.
Although the robotic arm has a force-torque sensor on its wrist, this sensor is not used in the collision-detection system; it only serves as a measurement reference in the experiment. “The collision detection and reaction,” Haddadin told me, “is based on a very good dynamics model of the robot and the fact that, unlike other robots, we have torque sensors and position sensors in every joint.”
With the dynamics model (which includes rigid body dynamics, joint elasticity, and motor behavior) and the sensor measurements, the robot can detect a collision nearly instantaneously. (The control system relies on a “nonlinear disturbance observer.”)
“This method does not require any additional external sensors and only relies on the internal capabilities of the robot,” Haddadin says.
This is the first study to investigate soft-tissue injuries caused by robots and sharp instruments. Previous studies by the same researchers, as well as other groups, have focused on blunt collisions involving non-sharp surfaces.
The video below shows impact experiments using crash-test dummies and large industrial robots. Ouch.
Erico Guizzo is the Director of Digital Innovation at IEEE Spectrum, and cofounder of the IEEE Robots Guide, an award-winning interactive site about robotics. He oversees the operation, integration, and new feature development for all digital properties and platforms, including the Spectrum website, newsletters, CMS, editorial workflow systems, and analytics and AI tools. An IEEE Member, he is an electrical engineer by training and has a master’s degree in science writing from MIT.