Modular Self-Reconfigurable Robots (MSRRs) are a strange type of robot that eschews a fixed form. Imagine a bunch of Lego pieces that can arrange themselves into different shapes depending on their circumstance. If they want to reach up and grab something, they assemble themselves into an arm. If they want to move into a tunnel, they assemble themselves into a snake.
While we don’t have to worry about sentient Legos, yet, this transformer-esque capability is the core idea of MSRRs. They’re built from small units, called modules, that can rearrange themselves into different shapes to solve different tasks. Researchers from Cornell University recently made an MSRR breakthrough: they’ve developed an MSRR that can perceive its surroundings and autonomously make decisions about what shape to take in order to complete given high-level tasks.
Meet the MSRR
The researcher’s robot is comprised of wheeled cubic modules that attach to one another magnetically and communicate over Wi-Fi. The modules also connect to a sensor component that includes multiple cameras and a computer to collect and process data about the robot’s surroundings.
To reconfigure itself, the modules can detach from their neighbors and adjoin to any other open face. In this way, the robot can take 57 possible configurations. The Proboscis configuration, for example, gives the robot a long arm in front. The Scorpion configuration has a horizontal row in front with a perpendicular tail behind it. The Snake configuration lines up all the modules in one direction.
In addition to its 57 configurations, the researcher’s MSRR has a library of 97 actions including pick up, high reach, drive and drop. Both the configurations and behaviors were developed in part by an open-source community and design competition previously hosted by the researchers.
Putting It(self) All Together
Though the researcher’s MSRR has an extensive set of shapes and behaviors, its novelty lies in its autonomy. The robot can enter an unknown environment with a prescribed task and autonomously determine how best to accomplish that task. The researchers showed this with three demonstrations:
- The robot had to find all pink or green objects in the environment and deliver them to a zone designated with a blue square on the wall.
- The robot had to place a printed circuit board into a mailbox located at the top of a set of stairs marked with pink tape.
- The robot had to affix a postage stamp to the mailbox.
In order to accomplish these tasks, the robot had to reconfigure itself multiple times. To retrieve the pink object from a narrow opening, it changed to Proboscis configuration. To climb up the stairs, it changed to Snake configuration. To place the stamp on the box, it changed back to Proboscis.
Though the demonstrations weren’t flawless—it took the robot 24 attempts to climb the stairs, for example—all three were eventually successful. You can see for yourself in this video:
This research is paving the way for practical, adaptable robots that could eventually perform jobs involving changing terrain, such as cleaning up after an earthquake.
“Modular robots in general are just fascinating systems because you’re not restricted by one shape, so there’s a lot of flexibility,” said Hadas Kress-Gazit, principal investigator on the project. “The hardware is still in research stages, but if we had commercial modular robots, they would be very useful for anything where the environment changes significantly. The robot should adapt to its environment, as well.”
You can read the team’s full paper in Science Robotics.