A graduate student at UC Berkeley is one of the authors of a paper about a robot that uses cameras alone to guide it in the wild. They were able to get their robot to walk up stairs, climb on stones, and hop over gaps thanks to the help of cameras.
A general idea of what walking in a park or up and down stairs would be is what the four-legged robot is trained to think. There is a single camera in the front of the robot that guides its movement. Reinforcement learning is an artificial intelligence technique that can allow systems to improve through trial and error.
The team decided to remove the need for an internal map in order to make the robot stronger.
Jie Tan, a research scientist at Google who was not involved in the study, said it was difficult for a robot to translate a camera's raw data into a movement. The work is the first time he has seen a small and low-cost robot like that.
A researcher at the University of Washington who studies machine learning and robotic control was not involved in the research.
Akshara Rai is a research scientist at Facebook who works on machine learning.
Rai says that the work is a good step towards building perceptive legged robots.
Rai says that the team's work won't help the robot figure out where to go in advance. She says that navigation is important for robot deployment.
More work is needed before the robot dog will be able to prance around parks or fetch things in the house. While the robot may understand depth through its front camera, it cannot cope with situations such as slippery ground or tall grass, Tan says; it could step into puddles or get stuck in mud.