Hello builds a platform towards more home robots


The path to the domestic robot is strewn with pitfalls. The number of success stories he delivers can be counted on one hand. The reasons for this massive disconnect are nuanced and complex, just like the interiors of our homes. Twenty years after the arrival of the first Roomba, the robot vacuum cleaner is starting to seem like a coincidence – more the exception than the rule.

Aaron Edsinger, the former head of Google Robotics and now CEO of Hello Robot, isn’t trying to build the universal home robot — at least not now. The Stretch Robot (not to be confused with the Boston Dynamics truck unpacking robot of the same name) is a platform on which the company hopes the next generation of home robots will be built. Watching it navigate a house in the demo videos is reminiscent of Nvidia’s line of reference robots.

The recently announced Stretch 3 is a robot with a wheeled base and a height-adjustable gripper. In the promotional video, you’ll see a few Stretches walking around a house, making beds and unloading the dishwasher – exactly the way people have long dreamed of a home robot.

There are, however, two very important caveats. The first is the price of $24,950. As someone who has been known to complain about high-end Roombas going over $1,000, it’s hard to imagine anyone paying the price for a new low-end car, especially given the shortcomings of the system for consumers.

This brings us to point number two: the system is controlled by teleoperation. There is of course nothing wrong with teleop per se. I have made this point many times. But a one-on-one human-to-robot control scenario isn’t sustainable, especially in the home, which you probably don’t want to open up to anyone on the other side of the camera.

One area where teleoperation is great is in the robotic learning process. This is where reinforcement learning comes in: guiding the robot through the process of performing tasks in different scenarios. That’s the kind of thing Tesla is probably doing in this recent Optimus folding laundry video – even though the company didn’t initially seem particularly keen to disclose these informations.

“Too often, a video offers an exciting glimpse into the future, but the robot is not available,” says co-founder Charlie Kemp in a statement. “Stretch 3 is not vaporware. It is available today. It’s an invitation to join an incredible community that is creating an inspiring future. It’s also the most fun I’ve ever had as a programmer.

All of this is true – except perhaps for the last part. We’ll just have to take the good doctor’s word on this. But being on sale today doesn’t mean most people will – or should – buy it. Much like the Nvidia example above, it’s rightly considered a reference device that third-party developers can access to create the kind of apps that might – one day – be genuinely useful.

Let’s return to the question posed at the beginning. Why have we waited so long for a true Roomba sequel? This product was designed to do one thing competently and has gotten much better at this single task over time. The initial Roomba had a hockey puck design and honestly didn’t stray too far from the first generation on that front. There are extreme limitations to this form factor, however, including height (this matters a lot when it comes to the placement of mounted sensors) and lack of limbs.

Image credits: Hello robotics

As for this second part, Hello tellingly references the recent enthusiasm around humanoid robots. The notion of “general use” comes up often. Remember, for example, when Tesla Bot was first announced and the company’s CEO promised a robot that could work all day in the car factory and then run some errands for you on the way? return ?

It would take many more words than I currently have to explain why truly generalized robots are much further away than you probably think. I have often discussed a happy medium between the two: moving from single-purpose robots to multi-purpose robots. The way forward could indeed involve an SDK and an app store type approach to introducing new features.

In this case, we begin to ask the reasonable question of how much the next domestic robot should resemble us? The really compelling argument here is for stairs, but we are far from the point where such mechatronic complexities can be delivered to home users at reasonable rates.

I find this excerpt from Hello’s press materials particularly interesting: “Hello Robot has paved the way for a middle ground between simple, single-use robots and complex humanoid robots, showing that robots do not need to be humanoid to perform a wide variety of demanding tasks. in the houses. »

Mobile manipulation constitutes a huge bottleneck in the development of a real domestic robot. It’s likely that the solution isn’t just a few arms stuck on a Roomba. Rather than jumping straight to building another robot in our image, Stretch offers a manipulator more in line with what I’ve seen in domestic robot research projects like those at the Toyota Research Institute.

I’d say, at the very least, that it’s a space worth keeping an eye on, even if you’ll have to continue to wait patiently for your next robot buddy.


Leave a Comment

Your email address will not be published. Required fields are marked *