The road to home robotics is fraught with danger. The number of success stories it has delivered can be counted on one hand. The reasons for this massive disconnect are as subtle and complex as the interior of your home. Twenty years after the first Roomba was released, robot vacuums are starting to feel like an accident. It’s more of an exception than a rule.
Aaron Edsinger, former Google Robotics executive and current CEO of Hello Robot, isn’t attempting to create a general-purpose home robot, at least not for now. The Stretch Robot line (not to be confused with the Boston Dynamics truck unpacking robot of the same name) is the platform on which the company hopes to build the next generation of home robots. Seeing it cruising around the house in the demo video reminds me of Nvidia’s line of reference robots.
The newly announced Stretch 3 is a robot with a wheeled base and height-adjustable gripper. In the promo video, the two Stretches can be seen running around the house, making the bed and unloading the dishwasher. This is exactly the kind of home robot people have been dreaming of for a long time.
However, there are two very important caveats. First is the price tag of $24,950. As someone who’s been known to complain about high-end Roombas costing over $1,000, it’s hard to imagine anyone paying for a low-end new car, especially considering the shortcomings of the system for consumers.
The second point is that the system is controlled by remote communication. Of course, there is nothing wrong with the teleoff itself. I’ve mentioned that several times. However, a one-to-one control scenario between humans and robots is not a sustainable scenario. Especially at home, you don’t want to reveal this to the person on the other side of the camera.
One place where Teleop excels is in its robotics learning process. This is where reinforcement learning comes into play. The goal is to guide the robot through performing tasks in various scenarios. This is most likely what Tesla is doing in its recent video of Optimus folding laundry. Although the company didn’t seem particularly keen to reveal that information at first.
“Videos give us an exciting look into the future, but there are too many instances where robots aren’t an option,” co-founder Charlie Kemp said in a press release. “Stretch 3 is not vaporware. Available starting today. This is an invitation to join an amazing community creating an inspiring future. It’s also the most fun I’ve ever had as a programmer.”
All of that is true. Except maybe the last part. We’ll just have to take the good doctor’s word for it. But just because something is on sale today doesn’t mean most people will or should buy it. Similar to the Nvidia example above, it’s best to view it as a reference device that third-party developers can access to create the kinds of apps that might one day be really useful.
Going back to the question I asked at the beginning. Why have we waited so long for a proper follow-up to Roomba? The product is designed to do one thing competently, and over time it gets much better at that single task. The original Roomba had a hockey puck design and honestly its front wasn’t that different from the first generation. However, that form factor has extreme limitations, including height (which is very important in relation to where mounted sensors are placed) and lack of limbs.
Image Credits: hello robotics
As far as the second part is concerned, Hello is referring to the recent excitement about humanoid robots. The concept of “general purpose” comes up a lot. For example, remember when the Tesla Bot was first announced and the company’s CEO promised a robot that could work all day at the car factory and bring you groceries on the way home?
To explain why truly generalized robots are much further away than you think would require far more words than are currently allocated here. I have often discussed the middle ground between the two: moving from single-purpose robots to multi-purpose robots. In fact, that path may involve SDKs and an app store-style approach to introducing new features.
In this case, a reasonable question begins to arise about how much the next household robots will have to resemble us. The really compelling argument here is the stairs, but we are far from the point where such mechatronic complexity can be delivered to home users at a reasonable price.
I find the following part of Hello’s press release particularly interesting: “Hello Robot blazed a middle path between simple single-purpose robots and complex humanoid robots, showing that robots don’t have to be humanoid to perform a variety of fascinating tasks. “At home.”
Mobile manipulation poses a formidable obstacle to the development of suitable home robots. Perhaps the solution is more than just strapping two arms to the Roomba. Rather than just building another robot in the image, Stretch provides manipulators that are more consistent with what we’ve seen in home robotics research projects, like the one found at the Toyota Research Institute.
I’d say this is at least a space worth watching, even if you have to continue to wait patiently for your next robot friend.