TidyBot: The Future of Home Cleaning Robots

TidyBot revolutionizes home cleaning by autonomously picking up and organizing messes effectively.

### The Future of Home Robotics: TidyBot and Mobile Aloha Innovations
* 🤖 Introducing TidyBot, a robot designed to **clean up messes** by identifying and picking up items, showcasing the potential for **general-purpose home robots**.
* 📸 Equipped with ceiling cameras and its own cameras, TidyBot can **locate and categorize objects**, making decisions on where to place them, like **tossing a Lego block**.
* 💡 Utilizing a large language model, TidyBot learns from user input, adapting to **where items belong**, and even has the capability to **make your bed**.
* 💰 However, the current cost of TidyBot is high, with the arm alone priced at **$40,000**, but potential exists for **cheaper consumer versions** in the future.
* 🍳 Meanwhile, Mobile Aloha is another innovative robot that can perform various kitchen tasks, including cooking, by **learning through imitation of human actions**.
* 🛠 Researchers aim to refine Mobile Aloha's design for home use, focusing on **reducing its size** and **enhancing its capabilities to operate independently**.
* 👥 The research at Stanford Robotics Center emphasizes not just task completion but also the need for robots to **understand and navigate human movement safely** in homes.
* 🌍 Ultimately, the goal is to develop robots that assist people, particularly the aging population, by helping them **live independently and safely** in their own homes.

Forget humanoid robots. What you really need is a robot to clean up all of your mess. Say hello to TidyBot, a rolling robot with a gripper arm that sees your mess, picks it up, and knows exactly where it needs to go: into the bin, or a drawer, or on the shelf. TidyBot is just one of many projects that pave the way for general-purpose home robots. I'm here at the Stanford Robotics Center with your first look inside. It's where students and researchers are working on the future of human-robot interaction. Quadrupeds, cooking robots, and even underwater MBots are all here. But let's get back to TidyBot. These cameras on the ceiling locate items and obstacles in your space, while cameras on the robot itself identify objects it picks up. If you have something like a Lego block, for example, lying around on your floor, it can also decide, um, if you told it so that this is actually a tossable item that won't break, uh, and uh, so that's of course much faster than uh, rolling up to everything and placing

it carefully. Heidi bot uses a large language model to categorize your stuff. After you show it a few examples of where things belong, it learns where to put similar items. And guess what? It can even make your bed. But we're still a ways off from seeing this Roomba on Wheels roll into our homes. Unfortunately, the tidybot itself, as you see it right now here, is extremely expensive. Just the arm costs $40,000, by the way. But if there was a consumer product company that would take this idea and use hardware that was actually much cheaper than what we are using, then I think the price can go, of course, down by a lot. If we think about the software, um, there I think we are getting definitely close to having a product like tidy bot actually driving around, um, in your apartment and picking up the right things. Autonomous cars already have this technology to just, like, navigate around spaces, so that could be something that is implemented on tidybot already. And we have ways to recognize new objects now and categorize them.

according to your preferences as well not to be outdone on cleaning mobile Aloha puts away pots and pans wipes down the counter loads the dishwasher and can even cook a simple meal autonomously. Welcome to Lexi's kitchen on the menu today delicious sautéed shrimp. Mobile Aloha learns how to do all this through imitation so I'm going to tell it to operate the robot while it captures images motion and manipulation data. It's like exactly reflecting the actions that I'm doing. Yeah you can move more drastically we kind of set the task whether it's sauté shrimp or store a pot into a cabinet. We randomize the position of that object or even the background of the scene and also maybe the different types of P you want to place it in and then repeat that process for 5100 time and use that data to train a model that will eventually control the robot. So I'm not going to train this 50 times but I am going to try to sauté a

shrimp and help train this robot. You can move around to feel the resistance of the robot. Oh, and turn around. So, I've just got to think of this as an extension of myself right now. All right, let's get my arm just holding on barely to the plastic. Yes, I got one. One more try. Yes, I got them all. Yes, it's like a video game. My shrimp are in the pan. I can hear the sizzle. Does require a lot of concentration. Yes, time to give this a little bit of a stir. This is kind of wild that I'm able to do this, and it's learning from me. Aha, there we go. So, how did I go? Zaping, rate me out of 10, you think? Eight. Eight? Okay, eight. That's not bad. Dinner is served. Mmm, delicious. I don't want to eat this. So, what's next for mobile Aloha? Obviously, this is not ready to be in our homes. We're aiming to reduce the footprint or the size of it and also make it

you know it can hopefully it can come climb series and go outdoors in the future. We hope that we can leverage some Foundation models from the natural language processing and computer vision to help the robot to understand what is going on, what task it needs to complete without any help from humans. But household robots need to do more than just chores. Here at the center, students and researchers are also studying human movement so robots can be better equipped to safely roam around our homes. We have mobile robots like the ones you see here that are running around and can run around safely among people, but you'll notice what they don't do is touch people, right? The next generation of robots needs to help people live longer on their own, right? And so to help people live in their homes longer, you need to be able to help them move around the home. You need to be able to keep them from falling. We need robots that can do that, that can be gentle and yet strong, um, and safe, and no robots are like that today. That's that's

the Next Generation, and that's why we study dance. We need to understand how people move and how robots can move around them. A lot of people think, when they think robotics, they think about task-oriented robots. It's replacing humans. It's taking jobs. This is not about trying to just replace all humans so they don't have to work, right? It's actually about where we don't have enough people to do jobs, right? That we don't have enough people to take care of the next generation of aging. It's about trying to solve the next generation of problems. In our case, we talk about advancing robotics, making it so that we can do with robots what you couldn't do before. Thanks so much for watching beta test. Make sure to subscribe and let me know what other robots you want to see going forward. Sweet dreams. Good night.

These AI Robots Can Do Your Chores for You

CNET is the premier destination for explaining and navigating the changing world around us. Through news, product reviews, price comparisons, and more, we’ll tell you what’s new in the parts of modern life that increasingly matter more — from health, home, and money to tech, culture, and climate — and how to get the most out of them.

CNET

TidyBot is a rolling robot with a gripper arm designed to clean up messes by picking up items and placing them in the bin, drawer, or shelf.

The robot uses ceiling and onboard cameras to locate and identify objects, and it learns where to put similar items through a large language model.

While TidyBot's current version is expensive, there is potential for a more affordable consumer product in the future, leveraging similar technology used in autonomous cars.

Mobile Aloha: The Autonomous Cooking Robot

Mobile Aloha is a robot that autonomously cleans and cooks, learning through imitation.

The robot's training involves capturing data from the user's actions and repeating the process multiple times.

The goal is to reduce the robot's size and footprint for future home use.

Researchers are working on leveraging foundation models from natural language processing and computer vision to help robots understand tasks without human assistance.

The next generation of robots aims to safely navigate homes, assist with human movement, and prevent falls, ultimately helping people live independently for longer.

The goal is to create robots that are gentle, strong, and safe, addressing the current lack of such capabilities in existing robots.

The Purpose of Studying Dance in Robotics

Studying dance helps understand human movement and informs robot mobility.

The goal of robotics is not to replace humans, but to address labor shortages and solve future challenges.

The focus is on advancing robotics to achieve tasks that were previously impossible.