Researchers at Carnegie Mellon University’s School of Computer Science and the University of California, Berkeley, have designed a robotic system which it claims can support a low-cost and small-legged robot to climb and descend stairs.
The team behind the bot explained that the robot it tested could ascend stairs of nearly its own height, while also managing to scale rocky, slippery, uneven, steep and varied terrain. Some examples they pointed to included walking across gaps and scaling rocks – it can also reportedly operate in the dark.
“Empowering small robots to climb stairs and handle a variety of environments is crucial to developing robots that will be useful in people’s homes as well as search and rescue operations,” said Deepak Pathak, an assistant professor in the Robotics Institute.
“This system creates a robust and adaptable robot that could perform many everyday tasks.”
The robot was designed to adapt quickly and handle challenging terrain by relying on its vision and a small onboard computer.
The researchers report that they trained the robot with 4,000 clones of it in a simulator, which also stored motor skills learned during training. This approach did not require any hand-engineering of the robot’s movement, which the university says is a departure from traditional methods.
According to the team, most robotic systems use cameras to create a map of the surrounding environment, but that this can often lead to inaccuracies in the mapping process.
The system the team used reportedly bypassed the mapping and planning phases by directly routing the vision inputs to the control of the robot, allowing it to react to oncoming terrain quickly and move through it effectively.
Moreover, no mapping or planning and the use of machine learning for training means the robot itself can be low-cost, with the team saying the robot it was able to use was 25 times cheaper than available alternatives.
“If it slips on stairs, it can recover,” said Ananye Agarwal, an SCS PhD student in machine learning.
“It can go into unknown environments and adapt.
“This system uses vision and feedback from the body directly as input to output commands to the robot’s motors.
“This technique allows the system to be very robust in the real world. If it slips on stairs, it can recover.
“It can go into unknown environments and adapt.”
A member of the team explained that the design of the bot was inspired nature, and the fact that humans and animals use vision to move. They also cited that previous studies had shown that blind robots — i.e. without cameras — can conquer challenging terrain, but adding vision and relying on it greatly improves navigation.
Hind legs on four-legged animals were another muse for the design, researchers said. For example, when a cat moves through obstacles, its hind legs avoid the same items as its front legs without the benefit of a nearby set of eye. This system was thus designed to work in a similar way.
