Robots can learn to swim and navigate through a maze in just a few hours, according to a new study.
The research published in the journal Nature Robotics found that the robots could be programmed to swim as they explore the environment.
The robot’s neural network would then send commands to its body to navigate the environment by sensing its position.
“This was a first-of-its-kind demonstration of self-learning that was based on simple and intuitive algorithms,” said lead author James LeVay, a postdoctoral researcher at the University of Washington.
The researchers’ findings are significant because they suggest that robots could learn to navigate environments more quickly than humans, and this could pave the way for autonomous driving systems.
“The future of robots is much brighter,” said LeVays co-author Paul Dallenbach, a senior scientist at the Carnegie Mellon Robotics Institute in Pittsburgh.
“We are seeing that machines are becoming smarter, and they can do things that humans can’t do.”
For example, the researchers found that an autonomous robot could learn by looking at objects and learning to recognize and follow patterns.
The robot also could identify patterns in a scene.
LeVaya said the robots were able to navigate through their environment as well as through the environment of their own control.
The robots could also be programmed with different tasks and be trained to learn new tasks by watching video clips of different robots performing different tasks.
“There’s a lot of room for improvement in robots’ ability to learn,” LeVayan said.
The robots would also have to learn to use different tools, such as their feet.
LeVay and his co-authors are now working on how to design robotic arms that could be able to perform various tasks, including walking and swimming.
They hope to develop a system that can perform tasks like climbing a ladder and navigating a room.
The researchers are also looking at whether robots could navigate in different environments, and are exploring the use of artificial intelligence for that task.
The research was supported by the U.S. Army Research Office, DARPA, and the Defense Advanced Research Projects Agency.
Follow the leader.
Follow Recode.com for the latest in tech.