Connect with us


These robots use AI to learn how to clean your home.



Share this:

A robot suspended from the ceiling slowly expands the arms of a sponge inside an ordinary looking house before wiping a kitchen cleanly.

A robot suspended from the ceiling gradually expands arms of a sponge inside an ordinary looking home before carefully wiping a kitchen surface clean. Nearby, another robot gently cleans up a flat screen television, which makes it slop slightly.

The robots clean inside a mock home in Los Altos, California. The researchers of the institute are testing a variety of robotic technologies that can help achieve the dream of a home robot.

After looking at homes in Japan, which were often small and cluttery, the researchers realized they needed a creative solution. The ceiling has been used repeatedly, says Max Bajracharya, VP Robotics at TRI.

In another Toyota laboratory in Cambridge, Massachusetts, the robots practice picking cups and plates and reloading the dishwasher. To work reliably and without damages the robots rely on artificial intelligence algorithms that learn how to hold objects correctly. The robot also uses a TRI-developed engine.

Toyota does not have a timeframe for commercializing its prototypes, but is looking for an early entry into a potentially huge market.

The Japanese automaker initiated the Toyota Research Institute with a $1 billion investment in 2015 and bet that its expertise in automotive production could be turned into advanced robotics for homes and workplaces.

Secrets of Machine Learning? Human teaching teachers are being called:

Gill Pratt, chief executive of the institute, says that the sensors, computing and actuation technologies found in modern cars basically make them simple robots. He says: A car is just a robot that runs on roads outside. “The purpose of a car is to increase a person and in the same way we think of robots—as amplifiers of human activity.”

There is hope, particularly in Japan, that robotics in the coming decades will help to care for an aging population. Robots could also keep homes clean besides people who need physical help and possibly even keep them company.

There is also a forecast that recent progress in AI will accelerate in the advancement of robotics. Today, most industrial robots are still relatively dumb, seemingly watching repetitive motions. But this is changing with new sensing and planning capabilities as well as some use of machine learning.

The robotics industry is already of money, and significant advances could multiply the figures in the next decades. A growing number of startup companies hope to commercialize AI-powered robots for simple, repetitive tasks in fulfillment centers and retail stores. In addition, big companies such as Amazon and Google invest in research that combines AI and robotics.

Having a robot perform useful tasks within an ordinary home is nevertheless a huge challenge because it involves adapting to complex and unpredictable environments. Although recent advances remain the ability to manipulate any everyday object, it remains an unsolved problem in robotics.

As TRI – Researchers admit – it is not clear how welcome robots will be in homes. However, TRI also conducts research on human-robot interaction besides test new hardware. And earlier in this year it announced that it would test robotics and other technologies near Mount Fuji in Japan.

As the TRI project shows, the creation of robots that could help in the home depends on AI as well as computer simulation and the cloud.

TRI researchers create a virtual reality environment to teach robots to perform a task by manipulating its arms. The robot uses machine learning after many tries to find the best movement. TRI’s robots also practice tasks such as loading a dishwasher in a simulated environment, which give them many more challenges to learn from.

We are starting to make some progress on how we [see if the robots learn the behaviors and then share that behavior across, say a fleet of robots,” says Bajracharya.

More great WIRED stories!

Share this: