If robots work together with humans, they will have to learn to ask for help from them. So say researchers from the University of Washington who are looking for new ways to teach robots. In their opinion, crowdsourcing can help in this matter, which will reduce costs and speed up the process of learning how to perform various tasks.
Human habitat is not very friendly for robots. Machines prefer to work in a tightly controlled environment, where everything is predictable, and objects are relatively easily identifiable.
The real world is extremely complex, changeable and poses complex tasks for robots that are not always within their power. For example, it is not so difficult for a robot to clean plates in a cupboard, like loading a dishwasher. Learning a robot to properly use another machine can take a long time.
Usually, robots are taught by observing and imitating humans. This method works, but it requires a lot of time and data necessary for accurate modeling of tasks. Robot interaction with just one person is not enough here.
The researchers suggested training robots using the online community. As a result, the machine will be able to receive many examples of solving the problem from different people and choose the best of the proposed options. The study with the participation of Internet users confirmed the effectiveness of this method, writes resource GizMag.
The results of the study were presented at the international conference on robotics and automation IEEE 2014 in Hong Kong. Engineers promise to share their further success at the Human Computation and Crowdsourcing conference in Pittsburgh in November.
Obviously, robots are becoming part of our lives. One of the countries leading in the field of robotics is planning to hold the Olympic Games of robots in 2020.
The article is based on materials .
Comments
Post a Comment