From Roombas to Rosie – Engineering Domestic Robots

Rosie the Robot. (Image courtesy of Warner Bros. Television Distribution.)

Proving that our distaste for household chores knows no bounds, people have been dreaming about robots that can do our housework for more than half a century.

From Rosie the Robot, who served the Jetsons in the 1960s, to Bender in Futurama, domestic robots have long been a mainstay of science fiction. Until recently, such autonomous and interactive robots were confined to the realm of imagination.

The majority of domestic robots that have made their way into our homes so far are relatively “mindless” creations designed to perform a single function. The most ubiquitous example is the Roomba robotic vacuum cleaner, which has been patrolling floors around the world since 2002. Joining Roomba has been an array of one-hit wonders designed to do things like mop floors, fold our clothes, clean the cat litter box, mow our lawns and clean our swimming pools.

(iRobot's Braava jet 240. A mopping robot. (Image courtesy of iRobot.)

More recently, advancements in artificial intelligence (AI) have led to the creation of more intelligent domestic robots that have the capacity to do multiple tasks and learn as they go. But there are still plenty of engineering hurdles to overcome on the way to creating a true Rosie: a robot that can do all your household chores, take care of the kids and even crack a joke to cheer you up when you’re having a bad day.

A lot of progress has been made, and some pretty clever robots are on their way to our homes in the next few years, but there are plenty of engineering challenges remaining on the road to building a personal domestic robot.

Human-Machine Interactions – Developing People Skills

It’s hard enough for us to understand each other, so making a machine that can navigate the complexities of human interaction is no easy task. However, if you want a robot that can not only take directions but anticipate your needs as well, enhancing human-machine interactions is a necessity.

Joseph Lyons, human trust and interaction branch technical adviser for the USAF, is given a ball by the collaborative robot Baxter after instructions were given to the robot by a member of his team. (Image courtesy of U.S. Air Force/Gina Marie Giardina.)

The rise of brain-computer interfaces (BCIs) is granting us the ability to give instructions to machines in new ways, but the path forward for domestic robots calls for machines that can understand our needs and interact with us using natural language.

We can already control machines to some extent using voice commands, as evidenced by the speech recognition capacities of programs like Apple’s Siri and Amazon’s Alexa. But, as the IEEE points out, proper autonomous robots need to go one step further, to the point where they can understand the nuances of human behavior and establish meaningful connections the same way we do with each other.

In other words, domestic robots (at least the kind you’d be willing to trust with your kids) need to have empathy. Researchers are hard at work on natural language processing and human-machine interactions—with some interesting results already—but the technology still has a ways to go.

Navigating Human Environments

As researchers at Stanford University have pointed out, today’s robots perform best when doing repetitive jobs like grasping and moving objects. Moreover, controlled environments like factories are well-suited to robotic automation.

A 45 minute exposure of a Roomba cleaning a room. The blue circles indicate spot cleaning, the result of a blue light that actives when the robot detects a large amount of dirt. (Image courtesy of Chris Bartle.)

However, as anyone who has had young children can attest, a household environment tends to be about as far from controlled as you can get. Our homes involve far too many variables to preprogram a robot that can deal with them all. These includes people and possibly pets moving around in spaces that are optimized for humans, not robots.

Add to that the fact that the environment can change without notice—for example, when remodelling—and it’s clear that successful domestic robots will need to be highly adaptive. The Stanford paper breaks the challenge of navigating a human environment down into five categories: perception, learning, working with people, platform design and control.

Researchers and engineers around the world are currently working on projects designed to overcome each of these individual challenges, but the ultimate challenge lies in finding a way to integrate the approaches into functional systems that will work for robots operating in the real world.

Reducing Sensor Costs

Just as buying a home computer in the ‘80s wasn’t financially practical for most people, a key challenge to any up-and-coming modern technology is cost. Although the costs may have come down in the last few years, many of today’s robots still aren’t cheap.

One of the reasons for this is that in order to successfully navigate its environment, a robot needs a whole array of sensors that are currently expensive to manufacture. Micro-electrical mechanical systems (MEMS) technology has brought down the cost of inertial sensors in recent years, but other sensing technologies like LiDAR are still fairly expensive.

The Leddar M16 sensor. (Image courtesy of Leddartech.)

Bringing domestic robots into the average home means bringing those costs down. Fortunately, recent interest in autonomous vehicle technology has spurred electronics manufacturers to find ways to produce LiDAR systems as lower costs. Leddartech, for example, has developed a proprietary solid-state LiDAR technology to cope with navigating densely populated urban areas. This could help reduce the cost of LiDAR, but that alone may not be sufficient to make domestic robots a more affordable proposition.

Domestic Robots Today

While our very own Rosie the Robot might still be out of reach from both engineering and financial perspectives, autonomous helper robots are starting to become a reality.

Take Zenbo by ASUS, for example. Announced in spring of 2016 with a price tag of $599 USD, Zenbo is branded as “your smart little companion” by its creators.

The company provided little in the way of technical details about the robot when it was released, but according to the demonstration, it’s packing at least one camera, speakers, a microphone and some kind of wireless connectivity.

In terms of what it can do, the video accompanying the announcement suggests that it’s not much help in the realm of practical household chores. It’s essentially a “tablet on wheels” that can take pictures, recognize voice commands, answer questions and control smart home devices.

There are a whole host of other interactive robotic devices like Zenbo being launched, which seem mainly designed to manage the parts of our lives connected to devices.

Then there’s Aido, an interactive “social family robot” that’s designed to play with your kids, help with household chores, handle your schedule and keep your home connected and safe. The only kicker is that it’s still in development.

Palo Alto-based Ingen Dynamics announced Aido in early 2016 with the launch of an Indiegogo campaign, with the plan to start shipping by the end of the year. As of this writing, after a raising $850,000 through crowdfunding, the company has so far only revealed the core functionality of the cute little robot in a prototype video released for early backers and investors.

At a cost of $1,200 per unit, Aido is not cheap. But if it really can do everything its creators claim, from helping you do yoga to playing with your kids—all while gliding around your house quietly on a ball—some people might see it as worth the investment.

Regardless of which venture you think has the best chance of being the iPhone of domestic robots, you can bet that someone is going claim that title. The spread of automation isn’t going to stop at your doorstep, even though there are still many engineering challenges to overcome.

Whether it’s a matter of building a better human-machine interface, finding more efficient ways to navigate unpredictable environments, reducing the cost of sensors, or more fine-grained problems like developing the necessary actuators and end effectors, Rosie is undoubtedly on her way.

What do you see as the biggest engineering challenge for domestic robots?

Comment below.