Andra Keay surveys bipedal robots like Boston Dynamics' Atlas and explains why warehouses, not humanoid tasks, are the real commercial opportunity. She traces how tools like ROS and affordable platforms from Willow Garage enabled startups like Fetch and Rethink Robotics to reshape robotics.
Chapters
And most of all, they're affordable. OK, this is an old deck. This is the old style of robotics. Port handling, Australian. I'm surprised that the Los Angeles ports aren't already using automatic loaders. And has anyone seen Boston Dynamics videos? Yes. Right. Everybody knows Boston Dynamics videos. Have you seen the humanoids, the Atlas robot? Anybody seriously expect those to be commercially useful? Okay, yeah. What do you think that is the most likely task for one of the Atlas robots to do in our environment? Okay, I don't think you need bipedal for surgical. operations. I mean, I guess that's kind of the point. Atlas is really a testing ground. There are not many use cases for bipedal robots. And what we're starting to see is that if you look around, there's a company, Agility, who've got a robot, Cassie, and SRI International. And they've been working on the bipedal legs so that they operate— like Atlas only recently could operate for You know, a few hours without being plugged in or attached to a tether. The SRI robot was able to get up to, you know, like 18 hours on a charge. I believe that Cassie operates a lot more. But the focus on legs, why do you want legs on a robot? To traverse uneven terrain. Okay? And we're getting better at building adaptively balancing robots based on biomimicry, based on on learning how spring is held in muscles and joints, learning how to create similar systems to animal systems. OK? But their most likely commercial application, if I can just speak in general terms, would be the legs shifting large boxes and cargoes around. Which is what this one was designed for, for army purposes. It's a, you know, stuff carrier. Okay, so we might see bipedal robots in warehouses moving stuff around for us. And we might see the top half of robots, or at least the robot arms, performing surgery with sensors and with teleoperation. And that's like the intuitive systems, which have been around for almost 20 years now. We're getting a lot more affordable systems like that these days. One of the things is there are very, very few reasons to put the legs and the arms and the head together yet, because we're not good at the level of behaviors that we expect from humans. And oh great, that's the intuitive Da Vinci surgical system. So really the human is providing the eyes and the arms, and it's being translated into multiple arm extensions. And it's wheeled, so you move it where it needs to be. It doesn't walk anywhere. PR2, it's really discontinued by the way. You'll still see videos of it folding towels at about 1 towel every 3 minutes. You will get newspaper headlines, robot folds towels. Yeah, no. No, no, no. PR2 is a research robot. The idea was to create a general platform hardware robot for the home use. Really, people worked out that we were still a whole generation of robots, or several generations of robots off that. Well off the price point. Well off the ability to do any of the jobs in the home. But there are 11 PR2s in the world today, and they're serving a good research purpose because they're the same hardware. One of the problems with robotics research is that it's not a science. Nothing is replicable. Everybody's built their own systems. Nobody else can get the same results without trying to build, even building the same system. PR2 was one way to start to change that around and create replicable results. The other great thing, this was done by Willow Garage. Anyone heard of Willow Garage? This is what brought me over from Australia, actually, Willow Garage. It was a beacon in robotics. It was founded with the personal funding of an ex-very early stage Googler who invested his fortune into trying to build a home robot. And he created a kick-ass robotics research and development group here in a garage, Willow Garage. And they had a lot of researchers in human-robot interaction and design. And they openly published all of their papers. So I mean, they really shook up the robotics world. It was fantastic. And what they also did, as well as building this giant PR2, which cost $375,000 apiece, which is why there's only 11 of them in the world. They built ROS, the Robot Operating System. And over the last 5 years, ROS has gone from being used by students and a few startups in Silicon Valley to being used in the prototyping stages of every single startup that I work with these days. You might not use it in your commercial release, but it fast-tracks forward your prototyping and development. And as more people use ROS, more people know ROS, more support is given to the software. This Willow Garage collapsed, but they spun off 8 to 10 companies that are really successful and interesting, ranging from computer vision to the Open Source Robotics Foundation doing ROS, which are getting a lot of commercial funding to develop ROS packages suitable for commercial applications, to Industrial Perception Inc. that was doing vision for industrial robotics and warehouse logistics, to Savioke who were doing the hotel delivery robot, to a lot of people that ended up in places like Bosch and Google and Apple doing cool things, to groups like Fetch Robotics And has anyone heard of Fetch Robotics? Okay, they're doing warehouse automation. And if I've got a photo of their robot, their founders were on the team that built this. And then they tried to build the cheapest possible robot using ROS that they could, and they went, they built the TurtleBot. Has anyone seen a TurtleBot or heard of them? No. Basically, it's a Roomba Create base, like a hackable Roomba base, with a netbook and a Kinect sensor. That was the first versions of it. So you could purchase one for around $1,400 with all of the hardware. And you could do remote operation. It could do mapping and autonomous navigation. It was operating on ROS. And that just was the most affordable research robot that had serious functionality that anyone had done. So that team went from building this Most sophisticated robot in the world to building the cheapest robot in the world. Rethink Robotics Baxter. We'll just skip on along. Universal arms, compliant and cheap. They've disrupted the entire industrial robotics industry, and they were just acquired by Teradyne or Stryker. I forget.
