Lakshmi Nair with the ‘MacGyver’ robot. Image: Georgia Institute of Technology
A robot that can create its own tools to complete simple tasks is being celebrated as a pioneer of a new era in robotics.
If robots are going to be sent to the most hazardous environments to potentially save lives or explore new worlds, they’re going to have to be creative. To that end, researchers at the Georgia Institute of Technology have revealed new tech that enables robots to build simple tools.
In doing so, it suggests robots might be on the verge of their own ‘Stone Age’, according to the scientists from the university’s Robot Autonomy and Interactive Learning Lab. This ‘MacGyvering’ technique – named after the famous TV show detective who could make escape tools out of seemingly anything – has been replicated using machine learning.
In a study presented at the 2019 Robotics: Science and Systems conference, the researchers gave a robot a set of optional parts and told it to make a specific tool. Trained with the right skills, the robot examines the shapes of each part and how they might be attached to one another.
For example, by learning that the concavity of bowls enables them to hold liquids, it makes use of this knowledge when constructing a spoon. In this study, the researchers were able to get the robot to successfully create hammers, spatulas, scoops, squeegees and screwdrivers.
In the example of the screwdriver, the researchers were interested to see the robot using the combination of a pliers and a coin.
“It reasoned that the pliers were able to grasp something and said that the coin sort of matched the head of a screwdriver,” said Lakshmi Nair, a PhD student working on the project. “Put them together and it creates an effective tool.”
So far, the robot is limited to the shape and the attachment, meaning it doesn’t have the knowledge of material properties to take the next, crucial step of applying them to real-world scenarios.
“People reason that hammers are sturdy and strong, so you wouldn’t make a hammer out of foam blocks,” Nair added. “We want to reach that level of reasoning in our work, which is something we’re working on now.”
