2024開(kāi)年，人工智能機(jī)器人迎來(lái)新突破，可以“包攬”家務(wù)活了。

 

 

 

 

近日，斯坦福大學(xué)IRIS實(shí)驗(yàn)室團(tuán)隊(duì)研發(fā)的機(jī)器人Mobile ALOHA亮相。

 

 

和以往更多用于工業(yè)領(lǐng)域的機(jī)器人不同，Mobile ALOHA將移動(dòng)性和靈巧性集成到雙手移動(dòng)操作中，澆花、煮咖啡、收拾衣服、打掃衛(wèi)生、幫人刮胡子、逗貓……這些家務(wù)瑣事它都能完成。

 

圖片

 

在研究者的訓(xùn)練下，它還成功完成了更復(fù)雜的任務(wù)——烹飪?nèi)�道粵菜（Cantonese meal）：蝦仁滑蛋、干貝燒雞、蠔油生菜。

 

圖片

斯坦福團(tuán)隊(duì)研究人員在社交媒體X發(fā)布的視頻截圖

 

 

打雞蛋、炒蝦仁、煎雞肉、切蒜末、切菜、灼水、加調(diào)味料，看起來(lái)“手”還挺巧……

 

圖片

圖片

 

Researchers at Stanford University recently announced a low-cost open-source hardware system for bimanual teleoperation, which can learn from humans to cook, clean and do laundry.

 

 

據(jù)團(tuán)隊(duì)介紹，該機(jī)器人Mobile ALOHA可以自主執(zhí)行任務(wù)，也可以由用戶(hù)通過(guò)遠(yuǎn)程操作系統(tǒng)控制操作。

 

 

研究人員在一篇論文中說(shuō)，Mobile ALOHA通過(guò)將原來(lái)的ALOHA——一種低成本、靈巧的雙手操作裝置——安裝在一個(gè)輪式底座上，擴(kuò)展了其功能，該基座能夠以每秒1.6米的速度移動(dòng)，與人類(lèi)的平均步行速度相似。

 

圖片

ALOHA構(gòu)造示意圖，圖源：研發(fā)人員@tonyzzhao社交媒體X發(fā)布

 

 

The robot, Mobile ALOHA, can perform autonomously or be operated by a user.

 

 

It extends the capabilities of the original ALOHA, a low-cost and dexterous bimanual puppeteering setup by mounting it on a wheeled base, which can move up to 1.6 meters per second, similar to average human walking speed, researchers said in a paper.

 

 

此外，該團(tuán)隊(duì)還為其設(shè)計(jì)了遠(yuǎn)程操作系統(tǒng)，可以同時(shí)操控機(jī)器人的底座和雙臂。該機(jī)器人重75公斤，能在1.5米的高度上施加100牛頓的拉力。

 

 

Moreover, a remote operation system has been designed, allowing simultaneous control of both the base and the two arms. With a weight of 75 kilograms, it can exert a pulling force of 100 Newton at a height of 1.5 meters.

 

 

在一段演示視頻中，研究人員展示了這款機(jī)器人如何自主完成烹飪蝦仁、呼叫并乘坐電梯、擦拭桌面上灑落的酒、將鍋放入櫥柜、清洗平底鍋、推椅子以及與人交互等任務(wù)。

 

圖片

斯坦福團(tuán)隊(duì)研發(fā)人員Zipeng Fu發(fā)布的社交媒體截圖

 

 

In a demonstration video, researchers showed that the robot can autonomously cook shrimp, call and take an elevator, wipe wine spilled on the desk, lift a pot to a cabinet, rinse pans, push chairs and interact with people.

 

 

另一段視頻中，通過(guò)遠(yuǎn)程控制，它還完成了洗衣服、使用咖啡機(jī)、自我充電、使用吸塵器、澆水、裝卸洗碗機(jī)、從冰箱取飲料并開(kāi)啤酒、扔垃圾、與寵物玩耍等任務(wù)。

 

圖片

斯坦福團(tuán)隊(duì)研發(fā)人員Zipeng Fu發(fā)布的社交媒體截圖

 

 

Additionally, through remote-controlled operations, it can do laundry, use a coffee machine, self-charge, use a vacuum cleaner, water plants, load and unload a dishwasher, obtain drinks from the fridge and open a beer, throw away trash, play with pets and more.

 

 

看到這里，你可能猜測(cè)，這款機(jī)器人一定價(jià)格不菲？

 

 

令網(wǎng)友驚訝的是，據(jù)團(tuán)隊(duì)介紹，ALOHA整個(gè)系統(tǒng)的成本價(jià)格只要3.2萬(wàn)美元（約合人民幣23萬(wàn)元）。它比現(xiàn)有的雙手機(jī)器人便宜得多，后者的價(jià)格高達(dá)20萬(wàn)美元。

 

 

不僅如此，研究團(tuán)隊(duì)已經(jīng)在將代碼開(kāi)源，即在網(wǎng)站上公開(kāi)了機(jī)器人所有軟硬件部分的源代碼，包括三維打印、組裝和軟件安裝等。

 

 

The entire system costs around $32,000. It is much cheaper than off-the-shelf bimanual robots, which can cost up to $200,000, according to an American technology website VentureBeat.

 

 

Researchers open-sourced all the software and hardware parts on the website, covering three-dimensional printing, assembly and software installation.

 

 

不過(guò)，這款機(jī)器人并非完美無(wú)缺。研究人員在論文中指出，無(wú)論是硬件還是軟件都需要進(jìn)一步優(yōu)化，目前仍存在一些“局限性”。

 

 

However, the robot is not perfect. Researchers said in its paper that both its hardware and software need to be optimized, as there are still "limitations."

 

 

研究團(tuán)隊(duì)成員、斯坦福大學(xué)博士生Tony Z. Zhao在一條推文中表示：“機(jī)器人還沒(méi)準(zhǔn)備好接管世界！”他還發(fā)布了一段視頻，展示了機(jī)器人在自主模式下犯的一些“愚蠢錯(cuò)誤”。

 

圖片

 

圖片

圖片圖片

 

"Robots are not ready to take over the world yet!" said Tony Z. Zhao, one of the three researchers, in a tweet on Saturday. He posted a video showcasing the "dumbest mistakes" the robot made in its autonomous mode.

 

 

據(jù)Tony Z. Zhao介紹，斯坦福團(tuán)隊(duì)接下來(lái)將對(duì)機(jī)器人進(jìn)行更多數(shù)據(jù)訓(xùn)練，以執(zhí)行更困難的任務(wù)，比如撿起和折疊皺巴巴的衣服。傳統(tǒng)上，洗衣服對(duì)機(jī)器人來(lái)說(shuō)非常困難，因?yàn)槲锲范逊e成它們難以理解的形狀。但他表示，團(tuán)隊(duì)的技術(shù)將幫助機(jī)器人解決人們以前認(rèn)為不可能的任務(wù)。

 

 

Next the Stanford team is going to train the robot on more data to do even harder tasks, such as picking up and folding crumpled laundry, says Tony Z. Zhao, a PhD student at Stanford who was part of the team. Laundry has traditionally been very hard for robots, because the objects are bunched up in shapes they struggle to understand. But Zhao says their technique will help the machines tackle tasks that people previously thought were impossible.

 

 

綜合來(lái)源：CGTN，觀(guān)察者網(wǎng)，果殼網(wǎng)，MIT Technology Review