What do the young, ultra-smart students at MIT’s Computer Science & Artificial Intelligence Laboratory (CSAIL) think about when they’re not hard at work? It turns out that it’s much the same stuff that similarly aged people everywhere think about: namely, the latest time-wasting viral challenge.
This month, that refers to the #BottleCapChallenge, in which a person films themselves removing the cap from a bottle in a jaw-dropping way (most commonly by spin kicking it). So far, it’s been attempted by an assortment of big names, including Conor McGregor, Jason Statham, Mariah Carey, and Jean-Claude Van Damme. Now MIT is getting in on the action by getting its Baxter humanoid robot to have a go. Using some neat mirroring technology, Baxter is able to copy the actions of a human operator who lifts their hand and, while doing so, adds the necessary spin needed to dislodge a bottle cap from the bottle it’s attached to.
“[We] made two important additions to the platform,” MIT CSAIL grad student Joseph DelPreto told Digital Trends. “We created our own soft gripper, which is made from flexible rubber and able to bend around objects like the bottle cap. It also has sensors that can help it determine an object’s shape to be able to better manipulate it. We created a system to control the robot, based on electromyography (EMG) sensors placed on a user’s biceps that monitor muscle activity in real time. We [then] developed algorithms that can continuously process these muscle signals to detect changes in the person’s arm level and allow the robot to mirror that motion or follow nonverbal commands. We used the Baxter robot for this task — but the algorithms to work with a robot using muscle signals could be used with any robot.”
While there are probably more sensible uses for cutting-edge robots than chiming in on the latest internet video challenge, Baxter’s attempt showcases the enormous potential for MIT’s real-time system for collaborative lifting.
“The system works toward creating more natural human-robot interaction during physical tasks by using muscle signals,” DelPreto continued. “We’ve tested it with a variety of team lifting activities, such as picking up objects and doing basic assembly. This could eventually be useful in factories or even around the home, allowing robots to be more effective assistants. As we add more wearable sensors or more learning capabilities, we hope to address more complex manipulation tasks. This #BottleCapChallenge was an interesting example that happened to also be something that lots of people are talking about on social media!”
- How emotion-tracking A.I. will change computing as we know it
- Move over, Spot. Anymal is a four-legged dog robot with all kinds of new tricks
- This drone with hands looks like a nightmare straight out of Black Mirror
- Japanese researchers use deep learning A.I. to get driftwood robots moving
- This robot arm for a wheelchair does everything from open doors to apply makeup