Computer Vision News - December 2022

9 Chelsea Finn that are high resolution, that don't break or change over time, and that are also fairly cheap. What prevents the hardware from becoming cheaper or more sensitive? The price of robots has been going down, and in many ways, that's because of motors. In the nicest robots, the motors are really kind of driving the cost of the robot. On some of the cheaper ones, there are these new motors called Dynamixel motors that are pretty nice and are a lot cheaper than the kind of motors that you find in industrial robots. But they also have their limitations. For example, a lot of the robots that we work with have parallel jaw grippers, which means they have two fingers open and close, kind of like a pincher. And the reason for that is if you want to build a more dexterous hand, like at these sorts of things. And if anything, I wish that we could have robots that have the sort of tactile sensing that people have. We have extremely high-resolution tactile sensing all over our bodies, especially on our hands. There are some tactile sensors out there that are starting to do better, but usually, they're either expensive, or they're somewhat brittle. So if you run lots of experiments, they'll degrade over time or they're very low resolution and they don't give you the very detailed feedback that humans can get. That's actually one of the reasons why for robots because we don't have these really awesome tactile sensors that humans have, we often actually just put cameras on the hand of the robot, and then you can get actually pretty nice sensory information in the same place that humans have sensory information. Is that what you call haptics? Yes, tactile sensing is kind of a form of haptics. But we often don't use haptics because the sensors are not quite there. Not yet. If I understand correctly, the main problem is the hardware, and the hardware is mostly not in your hands. If I look at the whiteboard behind you, most of the work that you do is work with formulas, software, and mathematics, and not with the hardware itself. If I could give you better hardware than what you have in stock, what would that be? The main thing that I love on the hardware side is the robots that we have right now, but ten to 100 times cheaper. That would be great! Because then we could buy lots more of them. I mean, we still already have probably more than ten robot arms in our lab, but we might be able to buy 100 if they were ten times cheaper. Then my second thing would be really good tactile sensors