Posted at 12:00 p.m.
Objective: to improve the daily lives of people
A robotic arm that picks up an abandoned stuffed animal on the floor. A second one who puts away the dishes and sets the table. A third who sucks up all the crumbs lost in the crevices of an armchair. Last May at the International Conference on Robotics and Automation, the company Dyson lifted the veil on one of its hitherto kept secret projects: the creation of an “autonomous device capable of carrying out chores household and other chores.
“There is a bright future in robotics,” says chief engineer Jake Dyson, in a video posted on the sidelines of the event.
“Allowing citizens to save time” and “improving people’s daily lives” are two objectives that drive the company, recognized for its vacuum cleaners, to further its research in robotics.
To carry out this project, which it hopes to see deployed in homes by the end of the decade, the company is currently conducting the largest recruitment campaign in its history. In one year, 2000 people have joined the team and hundreds of vacancies remain to be filled, especially in the United Kingdom.
Note that Toyota has also been working to create a robot that can do different tasks around the house for a few years.
The house, a complex environment
The idea of creating a robot capable of cleaning the entire residence is not new. As early as the 1960s, the character of Rosie, in the animated series The Jetsonsfed this fantasy.
“Making a robot physically capable of performing all kinds of tasks is very difficult, especially because our homes were built for humans,” says roboticist AJung Moon.
An example ? Door handles. They don’t all look the same and don’t all work the same way. While a human will quickly figure out how to open the door despite these differences, learning will be slower for robots, points out the assistant professor in the Department of Electrical and Computer Engineering at McGill University.
No two houses are the same. The number of rooms, the arrangement of the furniture, the way in which the objects are stored differ enormously. These variants complicate the creation of a robot designed to clean our residences. “We don’t have enough simulations of the typical house. […] We would need 10,000 examples of the variations that can be found in homes and the ways in which people organize their affairs,” says Glen Berseth, assistant professor in the department of computer science and operational research at the Université de Montréal.
Reproducing the human hand, an impossible mission?
The environment in which the robot operates is not the only challenge faced by engineers. “One of the reasons why we don’t have bipedal robots that clean houses is simply because it’s very difficult to reproduce the human hand, its ability to grasp objects, its skin, its way, for example, to take a glass without breaking it, explains Glen Berseth. There are very easy things that people do very well. You don’t have to think when grabbing a drink. […] The robot, however, must remember to be very careful. He is afraid of breaking the glass. […] When the robot has to perform gestures without being able to base itself on a very precise model and when there is an accumulation of uncertain variables, for example the shape of the glass, it becomes very difficult for it to perform its task effectively. »
Different prototypes
While conducting postdoctoral studies at the University of California, Berkeley, Glen Berseth and his team created a robot capable of picking up small objects. Thanks to artificial intelligence, more specifically reinforcement learning, the device, similar to a robot vacuum cleaner with an arm, learned by trial and error to perform this action.
At Carnegie Mellon University in Pittsburgh, Deepak Pathak and his colleagues used videos to teach their prototype to perform around 20 actions, such as opening a closet, folding a shirt or taking out the trash.
“Our approach is inspired by the way children learn. […] They improve in two ways: by practicing doing different things and by watching the adults around them,” explains the assistant professor.
Indeed, after seeing a human open a drawer in a video, the prototype failed to reproduce the action. A few hours of practice later, he was successful, says Deepak Pathak.
One robot, one task
According to experts interviewed by The Press, robots capable of performing many household tasks are unlikely to be found in homes for about ten years. Until then, we can turn to devices capable of performing a specific task.
Among the best-selling in the world is the iRobot Roomba. The American company has sold more than 40 million copies of its various robotic devices in some 70 countries.
The popular vacuum cleaner hit the market in 2002 and has come a long way since then.
While the first version wandered around the rooms somewhat haphazardly, today the device is guided by markers detected on the ceilings. “Light fixtures, door frames or window outlines are used as references by the robot so that it can determine where it is in the house,” explains Brent Hild, product manager at iRobot.
In addition to emptying its dust tank when it is full, the latest version of the robot also detects objects on the ground and avoids them. Roomba won’t drag the contents of a spilled bowl of cereal or a gift left behind by your pet, says Brent Hild.
At the mall
Robots haven’t just entered homes. They are also found in public places, such as shopping centers or airports. The only Canadian company to evolve in this niche, Avidbots created Neo, a driverless robot that washes floors in large commercial or industrial spaces. “Neo uses lasers and cameras to assess the world around him. The information it collects using its sensors is analyzed by its artificial intelligence. It decides which path the robot will take in order to maximize the floor area cleaned and minimize the time needed, without hitting anything,” explains Faizan Sheikh, co-founder of Avidbots.
Crossing a robot while shopping can be surprising. How do people react? “Most people want to take a picture with him. Especially the children,” replies the general manager of the company.
Questions
The growing place that robots could occupy in our lives in the coming years also raises its share of questions. “I believe we need to be careful how we introduce them into our homes,” AJung Moon said. The associate member of the Mila Institute stresses in particular that the data collected by them must be protected. Ethical issues will also be raised. “For example, if a teenage girl under the legal drinking age asks a robot to bring her a bottle of tequila, should the robot obey her? asks AJung Moon. “The problem is that you would have to plan for and program all these situations into the system. A rather complicated task, she believes.