Your office assistant! Alphabet restarts the robot project to challenge Ai puzzles

"At 1:30 in the afternoon, the robots will queue to the cafe. This group of monk robots will take four wheels to walk into the dining area, shuttle between the intricate dining table, then quickly launched, cleaning, and placing Table and chairs, everything is all settled in well. "

You can imagine that this is the scene that will appear soon? At present, Google’s parent company Alphabet is committed to bringing robots to the office to perform daily tasks. Their ultimate goal is to let the code "EveryDay Robot" robotAbility to complete more daily tasks, becoming a real office assistant!

Recently, the EveryDay Robots project, which is responsible for a team in its "X Lab", is committed to creating "General Learning Robots." They have removed some prototype machines out of the laboratory, enter Google’s Bay District Park, and implement some simple tasks.

"We are now running a team consisting of more than 100 robot prototypes, and they perform a wide range of useful tasks around our office," said Everyday Robots, Chief Robot, Bloom, said in a blog post. "The same robot of sorting garbage can now be equipped with a scraper to wipe the table, and use the same starting hand of the cup to learn to open the door."

The goal of these robots is to clean your hygiene. They will walk to the "meditation" for a few seconds before the table to determine if someone will hold. If someone eats, they will continue to advance until they find the empty table. After staying for 1 second, it may be in the internal operation, the robot is rotated and expanded, and the arm extends to the table, and there is a transparent disinfectant cover the surface.

This project wants the robot to become useful and can operate in an open environment, not limited to a controlled environment such as plants. After years of development, Everyday Robots finally sent to reality to do practical work.

The purpose of the R & D robber is to reliably complete the task of the human assistant. Because the computer is relatively easy to perform difficult cognitive work, it is extremely difficult to copy the cognitive function of the two-year-old child. For example, pass through a messy room, open a tricky door handle, and its difficulty is not overloaded.

These robots are essentially a wheel with a wheel, and there is a multi-purpose grip at the end of the flexible arm connected to the central tower. There is a "head" on the top of the tower, and there is a camera and sensor for machine visual, and there is a laser radar device that looks like a rotating. But even if the hardware is passing, the problem on the software is still unavoidable.

Even if this is, let Every Day Robots will open the door, and wipe the table and make people very crazy problems. In the context of current depth learning, engineers use machine learning to train software to identify objects and run millions of simulations to get a "almost" result.

The Chief Engineer of Everyday Robots said that some companies have now been very good at manufacturing machines that can make complex survivities in the factory, but when trying to copy simple tasks, such as cleaning up the kitchen or folding clothes, but fell into the kitchen or folding clothes. Dilemma, this is because cognitive problems have not been resolved.

In fact, now EveryDay Robots can do a major breakthrough in the timed fixed-point execution cleaning task. If Sophia can be so flexible in the future, it will definitely bring more help!