Facebook is trying to develop artificial intelligence models that will allow robots–including walking hexapods, articulated arms, and robotic hands fitted with tactile sensors–to learn by themselves, and to keep getting smarter as they encounter more and more tasks and situations. In the case of the spider-like hexapod ("Daisy") I saw walking around a patio at Facebook last week, the researchers give a goal to the robot and task the model with figuring out by trial and error how to get there. The goal can be as simple as just moving forward. In order to walk, the spider has to know a lot about its balance, location, and orientation in space. It gathers this information through the sensors on its legs.
"Much of our work in robotics is focused on self-supervised learning, in which systems learn directly from raw data so they can adapt to new tasks and new circumstances," a team of researchers from FAIR (Facebook AI Research) wrote in a blog post. "In robotics, we're advancing techniques such as model-based reinforcement learning (RL) to enable robots to teach themselves through trial and error using direct input from sensors." Specifically, the team has been trying to get a six-legged robot to teach itself to walk without any outside assistance. "Generally speaking, locomotion is a very difficult task in robotics and this is what it makes it very exciting from our perspective," Roberto Calandra, a FAIR researcher, told Engadget. "We have been able to design algorithms for AI and actually test them on a really challenging problem that we otherwise don't know how to solve."
As the term "machine learning" has heated up, interest in "robotics" (as expressed in Google Trends) has not altered much over the last three years. So how much of a place is there for machine learning in robotics? While only a portion of recent developments in robotics can be credited to developments and uses of machine learning, I've aimed to collect some of the more prominent applications together in this article, along with links and references. Before I delve into machine learning in robotics, go ahead and define "robot". Though at first this might seem simple, it's no easy task to come to an agreement on just what a robot is and what it is not, even amongst roboticists.
Facebook has publicly spoken about its interest in robotics in the past, but on Monday, the company finally shared details regarding the specific projects it's working on. The social media giant unveiled three robotics projects that it hopes will contribute to solving the ongoing challenge of building artificial intelligence systems that don't have to rely on large quantities of labeled data to learn new information. To do so, the company is conducting research aimed at teaching robots how to learn about the world, similar to the way that humans do. "The real world is messy, it's difficult," Roberto Calandra, a research scientist in Facebook's AI division said when speaking to Business Insider. "The world is not a perfect place; it's not neat. So the fact that we are trying to develop algorithms that work on real robots [will] help to create [AI] algorithms that, generally speaking, are going to be more reliable, more robust, and that are going to learn faster."