Many robots are great at performing singular tasks in a specific situation or environment, but they are not good at adapting in unknown environments to complete their objectives. Researchers today published a paper on a reconfiguring robot system that could carry out specific tasks by changing its physical design. Think Transformers, but without the explosions and collateral damage.
The modular self-reconfigurable robot (MSRR) system created by scientists at Cornell University and the University of Pennsylvania is composed of several modules that connect to one another to form larger robotic structures.
By reconfiguring the arrangement of these modules, the robots can transform from a wheeled robot into an arm able to lift and move objects, or a snake-like robot that can climb stairs. The results of their experiments is detailed in the report, “An integrated system for perception-driven autonomy with modular robots,” published today in Science Robotics.
“One of the big potentials of modular robotics in the autonomous field is that they can adapt their capabilities based on the environment they encounter,” said Jonathan Daudelin, a Ph.D. candidate at Cornell University, and one of the authors of the report. “You may have a robot that does really well at climbing stairs, but it may not be able to reach into a crevice. Or you may have a robot that can drive around like a car but can’t climb over uneven terrain.”
Tarik Tosun, another co-author and a Ph.D student in mechanical engineering and applied mechanics at the University of Pennsylvania, said a modular design adds flexibility as well as complexity because the reconfiguring robot has to make more decisions.
“At the very highest level of decision making, if we want to solve a problem with these robots, we have to decide not only what they’re going to do, like what you’d do for a regular robot, but also what they’re going to be — what shape they’re going to assume to complete their task,” he said.
“The question of autonomy becomes more complicated and more interesting when you have this kind of system,” said Hadas Kress-Gazit, one of the authors of the report and an associate professor at Cornell University. “The added complexity when we’re talking about modular robots is a design question, which is the shape of the robot.”
“We’re not restricted to one shape of a robot that can perform well in some environments, and maybe less well in other environments,” she said. “We can change the shape to actually be appropriate for the task it’s trying to do, and for the environment in which it’s working.”
The modular components are based on the SMORES-EP system, also developed at the University of Pennsylvania. Each cube-like module is about 3 in. wide, weighs about 1 lb., and has two wheels.
On four of the faces, an array of electro-permanent magnets allows any module to connect to another module. Because energy is only needed for the magnet when it makes or breaks a connection, the electro-permanent magnets can be beneficial for smaller robots with limited battery life, Tosun said.
Sensing the environment was done through a centralized sensor module, along with cameras and a computer, that made it easier for the system to make decisions for the higher-level tasks and reasoning for the overall objective of the cluster of smaller robots.
On the software side, the team used a suite of perception tools and library of configurations and actions for the reconfiguring robot to choose from. The system can reassemble itself into a scorpion, a snake, or a wheeled configuration to move around. The robot can also transform to a “proboscis” configuration that can pull and transport objects.
Reassembly does not even require that all the modules that remove themselves reconnect each time – in the video above, two of the mobile blocks stayed put while the snake-robot climbed “stairs” to deliver the object, reattaching themselves later.
Future applications for a reconfiguring robot could be as mundane as delivering letters or collecting objects around a home, or more importantly, in search-and-rescue scenarios.
“If you’re going into a disaster zone, it might not even be clear what the task is before you actually go in. If you’re going into a collapsed building, you don’t know what it looks like on the inside, or whether there are people in there that you might want to rescue,” said Tosun. “So having a robot that is very versatile can be useful in that scenario, because it could go in, assess its surroundings, and maybe choose to become a snake to go through a small crevice or even a shelter to protect people from falling rubble.”
Bonus video: Transformers!