- I'm Matei Ciocarlie. I'm an assistant professor in the mechanical engineering department at Columbia University. So we focus on dexterous robotic manipulation as an example of a complex motor skill, something that people do effortlessly but that robots have not yet acquired especially in complex, unstructured environments. When you don't know in advance exactly what kinds of objects you'll be confronted with, what kinds of scenarios and setups, that's when robots have tremendous difficulty in manipulating objects in the surroundings. You can think of multiple things, you can think of hardware, first of all. Maybe we don't have the right hardware. If you think about the human hand, such a marvelous, marvelous piece of engineering with 20 or even more independent joints depending on how you model them, and each of those joints driven by muscles with incredible ability to regulate and to control the forces being applied, so perhaps one thing that's missing is the hardware, that's a possibility. You can think about the computation, maybe we do not have the right computational ability even in terms of the mathematical models of manipulation or the ability to process the huge amount of data that manipulation entails, or maybe it's also a matter of just not having the right sensors. The human hand has amazing sensing capabilities which are often overlooked when you think about how flexible it is, and how strong it is, it's easy to forget just what an amazing sensing organ the human hand is, and it's possible that we just don't have the right sensing equipment to replicate that performance in robotics. And in our lab, we work on all three of those. We work on the sensing, the sensors for robot hands, we work on hardware and the design of fingers and joints and motors, but we also work at the computational level both trying to model manipulation from a mathematical perspective, but also to see if that our approach is that it can just learn how to manipulate things purely by doing, without necessarily modeling it mathematically at any point. One of the things that we really care about is tackling the hardware, the sensing, and the software together. And it's possible that one of the things that has held robotic manipulation back for a long time is the fact that you really have to think about the mind and the body together, and in academia, we often like to separate problems so we can focus on them individually, but in robotics, that is incredibly difficult because what is the body without the mind or the mind without the body in the context of robotics? So we care deeply about the hardware, the software, and the sensing together, and it's difficult to advance all of these at the same time but we strongly feel that you have to to get to the level where you can have complicated motor skills in the world. Robotics is the science of physically interacting with the world in intelligent fashion, so it's not just about the software and it's not just about the hardware. If you are to do that intelligently you have to do all of that. So, there is, there are many places for robotic manipulators. You can think of places where people do not want to go or cannot go, for example doing disaster response, emergency scenarios, or, for example, manipulating on board International Space Station where obviously it takes a lot of resources to get a person, to get an astronaut to do any task that needs done. So that is one step that you can imagine. Even here on Earth, there are many areas in logistics, in transportation of goods, where advanced robotic manipulation can have a lot of impact, and then finally, perhaps a bit counterintuitively, the most difficult environment of all for robotic manipulators might just be the home because just imagine the incredible range of things that happen in somebody's home, all the different types of things that you need to interact with on a daily basis, so tackling all that variation, that might just be the most difficult task for robotic manipulator, but I believe that eventually we'll get there too. It's, often people say that the ideal robotic manipulator will do chores around the house, and I believe we'll get there, but sooner than there we'll be in this type of semi-structured environments like the factory floor or the warehouse, or even the space station where the robot can still be very useful. It's a very interesting symbiosis in a way. When doing robotic manipulation, we try to learn as much as possible from the human hand and how humans manipulate things. Not copy the human hand, but just learn some of the underlying concepts, and this was from an engineering perspective a way for us to apply some of the technologies we developed and some of the things we've learned and to help people in the case where their own manipulation abilities are impaired. The unimpaired human hand is by far the most versatile manipulator known, but unfortunately, as a consequence of stroke, many, many people lose some of those incredible manipulation abilities. Almost a million people every year just in the US suffer a stroke, and many of them lose the ability to manipulate with one of their hands and arms. And when people lose that incredible ability that they have relied upon their entire life, all of the sudden very simple activities of daily living get very, very difficult. So this was a chance for us in a way to give back and use some of the robotic manipulation technology that we've developed to see if we can assist people who have lost the ability to manipulate on their own.