The path to the house robot is dangerous. On one side, it has provided success stories. This vast divide has multiple causes, including our houses. Twenty years after the original Roomba, the robot vacuum is becoming more of a rarity.
The former Google Robotics director and Hello Robot CEO, Aaron Edsinger, isn’t building the ubiquitous house robot yet. The business thinks the next generation of household robots will be developed on the Stretch Robot line, not the Boston Dynamics truck-unpacking robot. The demo movies show it cruising a house like Nvidia’s reference robots.
The new Stretch 3 robot has a mobile base and a changeable gripper. The commercial film shows two stretchers making beds and emptying the dishwasher—just what consumers want in a house robot.
Two major caveats apply. Starting at $24,950. As someone who has complained about high-end Roombas being over $1,000, it’s hard to envision someone buying a low-end new automobile given the system’s flaws.
Second, teleop controls the system. Teleop is fine in itself. I’ve said it many times. However, one-to-one human-robot control is unsustainable, especially in the home, where you definitely don’t want to expose anybody.
Teleop excels in robotic learning. Reinforcement learning guides the robot through varied tasks. Tesla undoubtedly accomplished this in the new video of Optimus folding laundry, even if the firm initially seemed reluctant to reveal it.
“Too often, a video offers an exciting glimpse of the future, but the robot isn’t available,” co-founder Charlie Kemp says via release. “Stretch 3 isn’t vaporware. This is accessible now. It invites you to join an incredible community that is building an inspirational future. It’s the greatest fun I’ve had programming.”
All of it is true, except maybe the final sentence. We must trust the excellent doctor on this. Being on sale today doesn’t guarantee most people will or should purchase it. Like the Nvidia example, it’s best seen as a reference device for third-party developers to create useful applications.
Return to the opening question. Why has a Roomba sequel taken so long? It was built to perform one thing well and has improved over time. The original Roomba had a hockey puck design and hasn’t changed much. Height (which affects sensor placement) and limblessness are severe restrictions on such a factor.
Hello’s second section references the current buzz regarding humanoid robots. The term “general purpose” is often used. Recall when Tesla Bot was launched and the CEO promised a robot that could work all day at car production and get groceries on the way home?
Explaining why completely universal robots are far off would require more words than I have here. I’ve long suggested switching from single- to multi-purpose robots. An SDK and app store-style method for adding new features may be necessary.
One can wonder how much the future house robot needs to look like humans. Stairs provide the strongest case, but mechatronic complexity is still too expensive for household users.
Hello’s press release: “Hello Robot has pioneered a middle way between simple single-purpose robots and complex humanoid robots, showing that robots don’t need to be humanoid to perform a wide variety of compelling tasks in homes.”
Mobile manipulation is a major obstacle to home robot development. More than two Roomba arms will likely solve the problem. Stretch provides a manipulator more like those from home robot research groups like the Toyota Research Institute, rather than developing another robot in our image.
I think this is a place worth observing, even if you have to wait for your next robot friend.
Hello is promoting home robots.
