For pretty much as long as robots have existed, humans have wanted robot butlers: autonomous machines that do our bidding around the home. But our imagination exceeds our technological capabilities, and the closest we’ve come to building Rosie the Robot is either specialized machines like robot vacuum cleaners or voice-activated gadgets like Amazon’s Alexa.

One Japanese startup thinks it can change this by dislodging a core tenet of home robotics: instead of having software control the machines, it wants humans to do it.

Founded in 2018, Mira Robotics unveiled its Ugo robot earlier this year, promising that the robot, when it goes on sale in 2020, will be able to carry out a variety of household tasks, starting with emptying a washing machine and folding clothes.

This may sound like a simple job, but it’s incredibly challenging for robots. Getting a machine to fold a T-shirt requires not only an understanding of how soft materials deform, but also the ability to perceive visual cues and the mechanical dexterity to manipulate that object. It’s knowledge that’s intuitive for humans but a real challenge to capture in code. Even cutting-edge research labs can’t make robots that can perform this task as nimbly as humans.

But by replacing robot brains with humans, Mira says it can sidestep these problems altogether. Speaking to The Verge, Mira CEO Ken Matsui says that if all goes to plan, robots that do your laundry will be just the start. “I believe remote-controlled robots can provide a variety of help in the home environment,” says Matsui, “like cleaning rooms, cooking preparation, a watching service for elderly people, caring for pets, or for tutoring.”

Ugo looks friendly enough to watch after your pets, but it probably couldn’t clean up after them. It has a wide base, a sensor-filled “head” with light-up eyes, and a pair of robotic arms attached to a cylindrical torso, which can telescope up and down like a firefighter’s ladder. Instead of humanlike hands, it’s equipped with pair of robotic pincers; they’re nimble enough to pick up clothes but not enough to unscrew a jar.

Mira is going to start beta-testing Ugo in “actual homes and actual families” this summer, says Matsui, before launching a paid service in 2020 targeting double-income families with children. Why this demographic? “Because they don’t have the time. They’re very busy, and we want to provide for them,” says Matsui. Prices have yet to be finalized, but customers can expect to rent the robot for around $225 a month, which will include six to eight hours of housekeeping services. Users can pay extra if they want more hours.

The company says it’s developing custom virtual reality motion controllers for operators, which will work with PCs and laptops via a custom app. Mira says it wants to employ these operators directly at first, but that, in the future, this work could be outsourced.

Ugo looks functional, but during demonstrations, its movements are slow and cautious.
Image: Mira Robotics

Operating robots remotely is not a new invention by any means, but to date, it’s a method that’s been limited to certain domains. Remote-controlled machines are usually found taking on tasks that are dangerous for humans, like bomb disposal or nuclear clean-up work, or they focus on transposing our social skills, not our mechanical ability, such as with telepresence robots used in offices or hospitals.

One big reason they’re not more common is that manually dexterous robots are incredibly expensive. There are suggestions that this could change, though. Earlier this year, for example, MIT unveiled a two-armed robot named Blue that costs just $5,000 to build. (That’s orders of magnitude cheaper than similar research robots that cost hundreds of thousands of dollars.)

AI could also help. Advances in machine learning mean some basic operations, like navigating around a house, can be done autonomously. In the case of Mira Robotics, the company will also use AI to blur faces and documents in the video feed seen by its teleoperators. That wouldn’t have been feasible just a few years ago.

Other companies are starting to explore new uses for teleoperation. Autonomous delivery robots usually have remote handlers, ready to take over if the bot gets stuck. Ditto robot security guards. San Francisco startup Kindred builds robot arms for warehouses. When its software fails, like when an arm can’t grasp an object, humans step in to operate the robot using VR motion controllers. This data is then fed back into the software to improve the robot’s abilities. Mira Robotics plans to introduce a similar feedback loop in its own service.

As Matsui says: “We believe our approach — developing low-cost robots and using AI along with humans — is the way to provide this service in peoples’ homes.”

There are plenty of reasons to be skeptical of Mira’s claims, though. In demo videos, Ugo the robot is slow-moving and clearly unable to match a human’s dexterity. It’s also bulky and unable to navigate stairs or even small steps. You could argue that these limitations wouldn’t necessarily matter in the right home, but they’re not promising attributes for the robot butler of the future.

Previous articleFacebook adds new digital birthday cards as part of its push into stories
Next articleOne year later, restaurants are still confused by Google Duplex