The robot butler dream doesn’t have legs

The idea of humanoid robots running our homes has been the stuff of science fiction for years. But unlike hoverboards and self-tying sneakers, a robot housekeeper has never felt close to rolling into our homes. Until now.

This has been the year of the humanoid robot. At CES, there were dozens of human-shaped bots on the show floor, a handful designed for homes. US-based robotics company 1X has already put its $20,000 Neo robot into homes to do housework (with a lot of remote assistance); and this week at the IFA tech show in Berlin, around half a dozen humanoids, mostly from China, will take to a catwalk to show off the alarming (?) acceleration of their skills as they plan to backflip into our lives.

But do we really need mechanical, multimodal robots that do our housework, or is there a smarter path that doesn’t involve 170-pound hunks of metal that could accidentally kill your dog with one misstep?

I’ve spent the last few years testing robots in my home; from robotic vacuums that can sweep and mop, to bots that mow my lawn, scoop cat poop, and cook dinner for me (all different bots, to be clear). I’ve also invited in Amazon’s Astro robot and several companion robots, including Ecovacs’ LilMilo, SwitchBot’s Kata, and Intuition Robotics’ ElliQ.

During my time with these robots, two things have become clear.

First, despite my tech-reviewer fascination with humanoid robots, a fully humanoid robot housekeeper isn’t practical. Unlike factories, where humanoids have seen some success, homes are messy, unpredictable environments. Also, most people have stairs. Solving that problem requires a robot that can either fly or have a pair of legs that can exert substantial force — neither of which is a good fit for a home. (Just ask Ring where its much-touted Always Home flying camera is).

Second, tech history has proven that precious few gadgets do multiple things well. We don’t need an expensive jack of all trades and master of none. At CES, I walked around the Las Vegas show floor with a bottle of laundry detergent, challenging every humanoid robot I met to open it. Not a single one could. I was repeatedly told it had to learn how to do it first.

What I think is more practical is a robot brain: an AI-powered smart home controller that can orchestrate the autonomous devices we already own based on the needs of the house and its occupants. Instead of one expensive machine trying to do everything, imagine a smart home that can proactively manage itself and adjust to the occupants: send the robot vacuum out when the dog spills his kibble, start the washing machine when energy is cheap, or dispatch the lawnmower when the edges need trimming.

Today, smart homes are mostly command-and-control or based on routines you create that require triggers. A robot brain could run your home autonomously and use reasoning to adapt when the home’s unpredictabilities arise. This appeared to be the idea behind Samsung’s Ballie robot (which never launched), and LG’s CLOiD humanoid (which looks more promising). It’s also the long-term vision of Amazon’s Alexa Plus and Google’s Gemini for Home smart assistants. No one is close.

At IFA this week, Tuya, a global AI cloud platform service provider, will launch a similar concept but with a tighter focus. Doova is an AI home companion designed to help seniors living alone. It is a short, round robot on wheels with no arms, but it does have “dynamic facial expressions.”

Tuya says Doova can track its owner in the house and be ready to assist in an emergency. It can open a live video call with family if they don’t respond. It will also provide “conversational companionship” and work as a smart home hub to control lights, curtains, and kitchen and bathroom appliances. The company hasn’t announced a price or release date, but says it will come to market.

Doova doesn’t need to fold laundry or load a dishwasher to be useful; it can understand what’s happening in the home, connect people to the home, and control machines already there. It does, however, present the other huge challenge for robots in our homes: They need cameras to navigate and interact, and cameras in our homes raise major privacy concerns.

Still, there is appeal in the idea of a robot brain that can do some manual chores — pick up clutter, load and unload a washing machine and dishwasher, and other tasks that require hands. I tested a robot vacuum with an arm that was supposed to pick up shoes and socks, and it mostly failed, but the idea is solid.

The middle ground between an imposing full-sized humanoid with legs and an armless robot companion is a wheeled bot with arms. That’s the direction LG took with CLOiD, a wheeled bot/smart home controller with arms boasting 7 degrees of freedom and five individually actuated fingers per hand.

I saw a live demo of CLOiD at CES this year using these appendages to perform simple tasks like folding clothes, taking items out of the fridge, and loading a washing machine, but not unscrewing a detergent bottle (I guess it uses pods?). Along with (very slow) physical actions, it also connected to and controlled smart appliances, including a washer, dryer, and oven.

This week, LG announced it is training hundreds of CLOiDs at a new data factory in Seoul, where they will practice a range of cleaning tasks in a home environment. By the end of this year, the company says this training data will feed roughly 100,000 hours of training data — equivalent to 12 years — into its own robot foundation model, built using Nvidia Cosmos open-world models.

It’s clear home robots are coming. But I don’t believe we need a robot butler that caters to our every need — very few people could afford one anyway. The future of home robotics isn’t one machine that can do everything, but a team of robotic machines, coordinated by a robot brain. And if that brain happens to come with a pair of hands that can open my laundry detergent, even better.