티디봇: 집안 청소 로봇의 미래

### 홈 로봇의 미래: 티디봇과 모바일 알로하 혁신 기술
* 🤖 물건을 식별하고 집어 올려 지저분한 곳을 청소하는 로봇인 **TidyBot**을 소개하며 범용 가정용 로봇의 가능성을 보여줬습니다.
* 📸 천장 카메라와 자체 카메라가 장착된 TidyBot은 **레고 블록을 던지는 것처럼 물건을 찾아 분류**하고 어디에 놓을지 결정할 수 있습니다.
* 💡 대규모 언어 모델을 활용하는 TidyBot은 사용자 입력을 통해 학습하여 **물건의 위치에 맞게 조정**하고 심지어 **침대를 정리하는 기능**까지 갖추고 있습니다.
* 💰 하지만 현재 TidyBot의 가격은 **팔만 4만 달러**로 높은 편이지만, 향후 **더 저렴한 소비자 버전**이 출시될 가능성이 있습니다.
* 🍳 한편, 모바일 알로하는 사람의 행동을 모방해 학습함으로써 **요리를 비롯한 다양한 주방 작업**을 수행할 수 있는 또 다른 혁신적인 로봇입니다.
* 🛠 연구원들은 모바일 알로하의 가정용 디자인을 개선하여 **크기를 줄이고 독립적으로 작동할 수 있는 기능**을 강화하는 데 중점을 두고 있습니다.
* 👥 스탠포드 로봇 센터의 연구는 작업 완료뿐만 아니라 **로봇이 가정에서 사람의 움직임을 이해**하고 안전하게 탐색할 수 있어야 한다는 점을 강조합니다.
* 🌍 궁극적으로는 사람, 특히 고령화 인구가 집에서 **독립적이고 안전하게 생활**할 수 있도록 돕는 로봇을 개발하는 것이 목표입니다.

TidyBot은 자율적으로 엉망진창을 효과적으로 치우고 정리하여 집안 청소에 혁신을 가져옵니다.

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

티디봇은 그리퍼 팔이 달린 롤링 로봇으로, 물건을 집어 쓰레기통이나 서랍, 선반에 넣어 어질러진 공간을 정리합니다.

이 로봇은 천장 및 온보드 카메라를 사용하여 물체의 위치를 파악하고 식별하며, 대규모 언어 모델을 통해 유사한 물품을 어디에 놓을지 학습합니다.

TidyBot의 현재 버전은 고가이지만, 향후에는 자율주행 자동차에 사용되는 유사한 기술을 활용하여 더 저렴한 소비자 제품이 출시될 가능성이 있습니다.

모바일 알로하: 자율 요리 로봇

모바일 알로하는 모방을 통해 학습하면서 자율적으로 청소하고 요리하는 로봇입니다.

로봇의 학습에는 사용자의 행동에서 데이터를 캡처하고 그 과정을 여러 번 반복하는 것이 포함됩니다.

향후 가정에서 사용할 수 있도록 로봇의 크기와 설치 공간을 줄이는 것이 목표입니다.

연구원들은 자연어 처리와 컴퓨터 비전의 기초 모델을 활용하여 로봇이 사람의 도움 없이도 작업을 이해할 수 있도록 돕고 있습니다.

차세대 로봇은 집을 안전하게 탐색하고, 사람의 이동을 돕고, 낙상을 방지하여 궁극적으로 사람들이 더 오래 독립적으로 생활할 수 있도록 돕는 것을 목표로 합니다.

목표는 부드럽고 강하며 안전한 로봇을 개발하여 기존 로봇의 부족한 기능을 보완하는 것입니다.

로봇 공학에서 무용을 공부하는 목적

무용을 공부하면 인간의 움직임을 이해하고 로봇의 이동성을 파악할 수 있습니다.

로봇공학의 목표는 인간을 대체하는 것이 아니라 노동력 부족을 해결하고 미래의 과제를 해결하는 것입니다.

이전에는 불가능했던 작업을 수행할 수 있도록 로봇 공학을 발전시키는 데 초점을 맞추고 있습니다.