It looks like a telescopic gripper tool attached to a hat stand, but it could be the future of home robotics. Meet Stretch, the first device from Hello Robot, a startup founded by former Google robotics director Aaron Edsinger and robotics professor Charlie Kemp of Georgia Tech, which came out of hiding after three years of development.
Stretch is not a consumer robot ready to roll into your home and start washing dishes, but a research platform that Edsinger and Kemp hope will build the foundation for home automation in the coming years. The bot's lightweight and affordable design could be a blueprint for future robots, especially those designed to take care of the elderly or people with physical disabilities. This allows companies to automate a range of household chores just as Roomba has automated vacuuming.
"Our long-term mission is to see that these types of robots are useful and helpful to people in households and at work," says Edsinger The edge. "What we're now expecting are research labs, corporate R&D, and venture-backed startups that use this for a variety of applications, all of which will drive the ball for these mobile manipulators."
(embed) https://www.youtube.com/watch?v=2msVU0ygrqM (/ embed)
The problem with these “mobile manipulators” – as robots with gripping arms and hands are known – is that they are large and expensive and represent a bottleneck for researchers.
Yes, some companies show humanoid robots at trade fairs claiming they can vacuum the floor or get you a beer out of the fridge, but these machines are just puppets that can only work under certain circumstances and are no more functional than Nintendo's ROBBING. Robot toy. At the other end of the spectrum are mobile manipulators used by scientists and researchers like Willow Garage PR2, are far more powerful, but also heavy, complex and expensive. For example, the PR2 weighs 220 kg and costs hundreds of thousands of dollars. Moving or testing in new environments is not easy, and is a significant investment.
"It's like a huge old computer when you want a laptop," says Edsinger.
In comparison, stretch was designed with simplicity and low cost in mind. Instead of using an expensive industrial arm with multiple joints and degrees of freedom, a single telescopic gripper is attached to a central hoist that achieves a similar range of motion at a fraction of the price. A single stretch unit costs $ 17,950 – still expensive compared to the consumer market, but a bargain for academics and researchers.
The arm is sensitive to touch for safety reasons and can carry objects weighing up to 1.5 kg. It's definitely a hit in terms of functionality (for example, it's not strong enough to open a fridge), but it goes well with moving things around in flat planes like tables, shelves, and counters.
Stretch has a wheel base like a Roomba with a size of 34 x 33 cm, making it small and nimble enough to maneuver in tight spaces such as corridors and small kitchens. And it uses a low-cost Intel RealSense depth detection camera and a 340-degree lidar sensor to monitor and navigate the environment so that it can be remotely or autonomously operated. There is also an integrated microphone array in the head that supports voice recognition for voice commands and a programmable LED light ring to give users feedback.
When a single design selection illustrates Hello Robot's philosophy, "Simple is better", it is the robot's gripper: a minimalist piece of hardware that consists of a pair of rubber shells and some metal springs.
Kemp says that when designing the bot, they quickly dismissed the idea of using human-like hands that are overly complex and can easily break. Instead, they wanted a grabber that could withstand real tests. That's why they turned to a pre-built research group: Amazon users reviewing home search tools for people with disabilities.
"We've gone through thousands of reviews, looked at the top rated grippers used by real people in real homes to grab objects they really wanted, and this (design) has changed in both the Amazon and Amazon reviews also prevailed in our laboratories, "says Kemp.
"We were able to create a robot version of it and it's really so versatile and forgiving. It doesn't have to be in the right place to work, it's just good at it to grab on things. "
While the hardware is polished, the unsolved problem for stretch (and more generally for home robots) is how to control these machines. Will they use AI to perform tasks autonomously, or will they be operated remotely, such as the mobile delivery robots that deliver groceries and takeaways?
Originally, according to Edsinger, the Stretch team introduced itself as a teleoperated consumer product that could help people in assisted living scenarios, for example. However, they soon found that the economy did not work. Stretch is relatively powerful, but like all mobile manipulators, it's incredibly slow. And if you hire someone to control a robot remotely for hours to do just a few simple tasks, it's cheaper to get them to do the job personally in a few minutes.
“When we got involved, it became clear that a certain degree of autonomy was required,” says Edsinger.
To this end, the company has provided Stretch with some basic autonomous functions. It can independently navigate through rooms and grab and hand over objects as soon as the action has been started. However, it cannot perform complex tasks such as folding clothes or cleaning surfaces itself. It is up to other researchers to test and implement them.
(embed) https://www.youtube.com/watch?v=kNNQd3cvNmg (/ embed)
However, the stretch & # 39; agnosticism reflects Edsinger and Kemp's wishes to ensure that the project has a long lifespan. As the mantra of the robotics industry is: robots are heavyStartups in this area often fail because of real challenges.
Edsinger is a veteran of Google's failed entry into robotics. The startup Meka Robotics, which he co-founded, was bought in 2013 together with the search giant a number of other robot companies, including Boston Dynamics. There were high hopes for Google's market entry, but the company's efforts failed when it recognized the limits of consumer robotics.
Even now, says Edsinger, "the business case for a consumer robot isn't quite there yet … the technology isn't quite there yet."
The answer, he says, is not to speed things up. He and Kemp avoided the “rocket fuel” of venture capital in the hope of building a sustainable business through sales. The research edition of the stretch robot has already been sold to half a dozen research laboratories, which they say provide invaluable feedback on what the market really needs.
"We realized that the best we can do to really solve this problem is to be there for a while," says Edsinger, "so as not to take a huge swing for the fences and disappear when we'll miss it. "
Hide and search with robots
With all the unanswered questions Hello Robot faces, Edsinger and Kemp are undoubtedly adamant about the possibilities of stretch. It's simply simpler and more user-friendly than anything else on the market, it says, and it opens up many opportunities for home robot research.
For example, Kemp has been using stretch prototypes in his home for years. He used the robots to play with his children and do basic tasks. On a Thanksgiving Day when he wasn't home with his family, he even used stretch to feed his cat. He controlled the robot with his laptop over the Internet and was able to open cans of cat food and put them in a bowl, fill glasses with water from the fridge and pour them out.
"I had this long-term dream of having a versatile robot in my house and now I have one, it's just a lot of fun," he says. "If you have a robot worth $ 400,000, you can hardly imagine hiding Easter eggs for your children with it. It restricts your thinking if you have these huge, expensive robots." With a machine that does less than $ 20,000, however, this is not the case.
Both men are understandably caged when it comes to proposing a time frame in which a robot like Stretch could do useful work as a consumer device. But it's certainly a question of years, maybe even decades. They point out that Roomba ("a phenomenal product") has taken decades of work to achieve its current capabilities and does only one, relatively simple task. They hope to develop machines that can do much, much more.