toggle menu
Episode 03

Mitchell Weiss has a long and distinguished career in the field of robotics with special attention to the field of manufacturing systems and mobile robots and is currently the COO of Piaggio Fast Forward. He is Co-author of Industrial Robotics: Technology, Programming and Applications (1986) and Instructor of automation and robotics courses at the University of Pennsylvania and Pennsylvania State University, and of engineering design at Massachusetts Institute of Technology. On this episode of the PFF Podcast, he provides an overview of what has changed and what hasn't over the course of three decades and calls special attention to the challenges presented by the development of robots, like PFF's own gita®, that operate in unstructured environments like towns and cities.

Transcript

Welcome to the Piaggio Fast Forward podcast. Join the conversation by subscribing to the PFF podcast at https://www.piaggiofastforward.com/podcast.

Jeffrey Schnapp

Welcome to the PFF podcast. I'm Jeffrey Schnapp, Chief Visionary Officer at Piaggio Fast Forward, and I have the pleasure of having a conversation today with Mitchell Weiss who is the Chief Operating Officer for Piaggio Fast Forward and a distinguished roboticist, with particular interests in the connection between mobility and robotics. Welcome, Mitch.

Mitch Weiss

Thank you, Jeffrey. Glad to be here. It is where I am every day.

Jeffrey Schnapp

And we do work together, so this can be a fun conversation. We can talk a little bit about your movement through different areas of robotics and how you got to PFF, and I guess a good starting point for that might be to ask you just how you got interested in robotics when you first started out?

Mitch Weiss

I was 15 years old, sitting in class one day, reading I, Robot by Isaac Asimov, and this would be a long time ago, folks, and one of my classmates walked in to the room and dropped on my desk a flyer for the Metro Toronto science fair, enter something in the science fair. And there I was, reading a book about robots and I said, "Oh, I know. I'll build a robot." So I did. Tragically, terrible robot, but I built a robot. And then I did it the next year and I did it the next year. And the second and third year I did it, I was actually winning pretty good prizes. So I decided I was going to go into robotics, and went off to MIT, which is where my computer science teacher thought I should go. And MIT had one class in robotics available, this is back in the '70s, so I had to negotiate with the institute to get a special do-it-yourself degree thing where I could spend time in mechanical and electrical and computer science so I could study robotics because there was no course of study back then.

Mitch Weiss

And got out of the school looking for a job in the robot industry and went to work for Unimation, which, for those who are great history aficionados, was the first robot company, started in the late '50s, early '60s. So that was my arc in.

Jeffrey Schnapp

Subsequent to that experience, you were Chief Technology Officer at Brooks Automation. Then you moved on for a number of years to serve both as Chief Operating Officer and Chief Technology Officer at the Seegrid Corporation before joining Piaggio Fast Forward in October of 2017. How has the field evolved over the course of those different roles that you've played?

Mitch Weiss

That's probably the most disappointing thing of the over 40 years I've spent in the robot industry. When I go to the industry shows and I see the products that are out there, they're not that far removed from what we were shipping in 1980 to what is shipping in the industry today. It's just beginning, and by just, last five, 10 years, to evolving to machines that are more autonomous. Up until now, robots were more automatic. They could repeat what they were told to do and told to do either through programming or what we called walkthrough then work, or lead through then work, but very deterministic kind of machines. Of course, even in the '80s, we touted there was AI involved and they were doing heavy math and making decisions on their own, but it was really very simplistic, deterministic stuff.

Mitch Weiss

So it's now, and I think, partly because of all the money that's gone towards the autonomous driving, really, if you look at the trigger, it's the DARPA Grand Challenge, and that was the kind of tech we were using at Seegrid, stereo vision navigation, that used a lot of what was called AI back then and is now called perception and pattern recognition. And we're barely starting to see the machines be truly autonomous. Even the self-driving cars are still very limited as to their routes, the actions they can take that aren't pre-programmed, predetermined actions.

Jeffrey Schnapp

Is this the reason you believe that while robots have had a transformative role in production and manufacturing and factory environments, that they haven't really entered the sort of everyday life world of human beings?

Mitch Weiss

That's correct. The factory environment is very structured. The next big growth area for robotics has been in what we call the distribution in order fulfillment space driven by the Amazons and Walmarts of the world, and that's the work that Seegrid started doing back in 2000. That's less structured than factories, but if you compare that to a New York City sidewalk as life-changing colors, that's a completely unstructured environment with a lot of random events happening. So the machines not only have to know what they're allowed to do at this instant, but because they're moving fast enough and in people's space enough, they really need to be predicting what people are going to do.

Mitch Weiss

If you think about how you drive down the street, you're continually scanning from side to side. If you remember what they taught you do in driving to school you are, you're scanning from side to side and you're trying to anticipate whether that guy's going to make a left turn in front of you or that person's going to step off a curve. You're not waiting for the event to get into some narrow field of view to take some immediate action. And autonomous cars still aren't predicting behaviors. They're still reacting to their environment. And unstructured environment, that's not a very smooth way to work with people.

Jeffrey Schnapp

I'm wondering if you could talk a little bit about the differences between the kind of work that you did as COO and CTO at Seegrid with what you're doing at Piaggio Fast Forward.

Mitch Weiss

At Seegrid, we were very interested in robots that did exactly what they were supposed to do for specific tasks. And one of the big challenges in that space would be retrieving a pallet of goods with a fork vehicle. So imagine an automated fork truck going to pick up a wooden pallet. It's an incredibly difficult problem, as much as we think it's a simple one, because pallets are almost always in the wrong place, almost always damaged, almost always unbalanced, and it's just a whole gamut of problems where you have to change the whole infrastructure from getting higher quality pallets, to better positioning, to all kinds of things, to make it work seamlessly with people. So typically, automation in industry, and even things like traffic lights, if you consider them to be automated device as everything else, require a whole rationalization of the workspace where you say, when this happens, then that happens and this always happens here and that has to be there.

Mitch Weiss

And the huge benefit that robots brought to industry was the rationalization in the workplace. If you had materials sitting in the aisle ways or robots building cars where the parts didn't fit together perfectly, nothing would come out of those factories. So once you go to automate them and the automation isn't as flexible and clever and adaptive as the humans who work there, all the components that come into the building and the structure of the building and the cleanliness of the building have to be perfect for the robots to work. Well, once you do that, the people work a whole heck of a lot better as well. So I always used to say one of the dirty secrets of the industrial robot business was you didn't really need to put the robots in to get the benefit, you needed to get ready to put the robots in to get huge benefits in your throughput and quality and everything else.

Mitch Weiss

What we're doing here at PFF is not looking to structure human lives, but really responding to how people work. So we've looked at gita as what it can do as a helper to humans, and then how would humans want it to behave in their space, on their terms. So rather than engineering the path that gita's going to follow to get through a doorway or to approach you as you want to put something in it or take something out of it, we've studied how people like machines to respond to them and we're trying to give gita that kind of behavior without making it a rigorous, pre-programmed, straight line, super-efficient motion.

Jeffrey Schnapp

How does that change the kind of engineering and technology development, software development work that has to be carried out in order to, in a sense, reverse the equation rather than trying to persuade humans to be more like robots per se, robots to respond to and be more like humans?

Mitch Weiss

Well, it's very different in some ways and exactly the same in others. What we do here is we studied the humans instead of the machines. So we look at the variables in how people move around and try to understand those as variables, not as fixed answers. So some people walk around a corner differently than other people do. So if you study enough of that, then you can turn gita into something that can slightly predict what the typical intention of the person is and respond tightly to the person, but not so tightly that it's unnerving. So if you're walking down the street, gita will follow you pretty much on the same path you're walking on. Not exactly. We won't go, if you're doing the drunken sailor walk, we won't be as drunk. We'll be a little less drunk, so it's smoother. When you go around a corner, we'll go around the corner on the same radius that you're going around the corner on, instead of making a beeline for your ankle and cutting the corner.

Mitch Weiss

And when you go through a doorway, we'll walk through the doorway at a respectful distance from your feet so you're not worried we're running over your toes. But for every person in every doorway, those distances are a little different. So we're looking at kind of comfort zones where we want the machine to land when it's doing something and designing those behaviors into the machine in terms of distances and accelerations and velocities that would make people feel comfortable. So that's always the measure of success is how does it compare to someone walking with a partner or a child or a dog? And when do they get uncomfortable? You remember the Seinfeld thing on close talkers? Well, you don't want gita to be a close follower or too distant that you're worried about it running away from you.

Jeffrey Schnapp

What you're describing is really the etiquette of moving around the world. And of course, that etiquette is something that we as humans, as bipeds, learned through really long and deep experience that starts in our infancy, so it's far from a simple proposition to transform that into something that makes sense from the standpoint of a robot. And also, mobility is of course a domain that has all kinds of cultural and ethical components to it. A lot of people struggle with mobility, whether it's because they can't carry backpacks or because of age or infirmity. What do you think about the kind of ethical dimensions of work in robotics? You were just describing some of the technical challenges, but the successful fulfillment of these technical challenges has real kind of social impact on how we design cities, how we think about our social interactions, the places we live, the people we share those places with.

Mitch Weiss

So there's two prongs to that, as you're asking the questions, two big topics for jumping around my head. One is the cultural piece of it. I fully expect how we want gita to behave will be different in different cultures. I noticed when I go back home to Canada once in a while and stand in line at a supermarket, actually, my wife noticed this because it made her so uncomfortable, people stand closer in Toronto than they do here because they come from a different cultural background. It's not Canadians, it's more, Toronto is a very cosmopolitan city. So, and when I think of Japan, I think of the time I was walking up the right side of a staircase and bumped right into somebody who was walking down the left side of the staircase because they walk on the left.

Mitch Weiss

So we're going to have to not only imbue gita with this responsiveness to the people in its area, but understand the social norms of where we're doing it. Mostly being closer, farther, left, right, kind of things, but well, I'm sure we'll find other stuff in other countries. And when you talk about the ability for robots to work with human beings to make human life better, the word I think of is accommodations. Robots are accommodators and the environment has to accommodate them as well. But my whole life in the robot industry, whether it was industrial robots or mobile robots or what we're doing with gita has always been about making life better. So in the industrial robot space, while there was always the stuff about you're replacing human jobs and everything else, robots got their start because they did the jobs that people couldn't do, didn't want to do or shouldn't do. We were carrying heavy things in dangerous places. The first robots I worked on were in nuclear hot cells. You don't want people in nuclear hot cells.

Mitch Weiss

Gita is going to try to do the same kinds of things for the public at large in these unstructured, etiquette-requiring outdoor and indoor environments, but I'm one of those guys who can't carry too much stuff. There's too many titanium rods in my back. And so for me, gita looks like a no-brainer, a backpack that I can carry and not have to carry. So as long as the cities are accommodating to that, gita's fine. The only way to get around unaccommodating spaces, and this is DARPA's new grand challenge, rescue robots going into disaster zones with lots of legs and things, it's not really a super-efficient, brilliant solution that you're going to scale out. We're not all going to have little horses walking beside us.

Mitch Weiss

So a couple of wheeled machine that can roll where we can walk, carry stuff for us, do the job for is without requiring us to do extra, hence it behaves as we would expect it to behave as a helper, is I think something we should be doing, something that in America, we're pretty able to do because the ADA, the Americans with Disabilities Act, has driven a lot of accommodation of space. There's still some challenges, but it's still a good place and something we should do. And we're seeing the same thing in the industrial robot space. All the buzz over the last few years has been in cobots, robots that can work beside you but when they hit you in the head, they don't give you a concussion. They're gentler. That's really the big delta, I think.

Mitch Weiss

And gita is designed to be kind of soft and cuddly. If you look at the shape of it, it's not soft like a teddy bear, but there's nothing gita can do to hurt you. It can only help you. So we think of robots as ways to make people's lives better, not to take away from the things that people do.

Jeffrey Schnapp

And along those lines, particularly with respect to robots entering into the everyday life world, sort of moving out of the factory, out of the warehouse, into the streets, into living rooms, the passage to consumer electronics hasn't really been a smooth or an enormous one. I'm wondering what your views are on the prospects for the increasing, weaving together of everyday life experiences with robotics.

Mitch Weiss

So in industry, robots have sold based on their return on investment. I can put a robot in so I can produce this many more cars an hour, I can do it with this much less labor. Yes, robots take jobs but transfer them to other places. In the consumer market, that's not there. The robot has to give you some intangible benefit and be something you can afford. You've probably heard this before from plenty of people in the robot industry and it's kind of stunning to me how many robots, startups there are every day now, although there were plenty in the '80s and '90s and aughts and '10s. Robots are really hard because it's the only technology out there.

Mitch Weiss

And when I say robots, I'm thinking of self-driving cars and self-flying airplanes and everything else where you have to take all this mechanical stuff and all this electrical stuff and all this software stuff and all this sensory awareness stuff. And now, we're adding human interaction stuff with it and make it work and do it in a timeframe that you can do it to develop a product and bring it to market and oh, by the way, make it cheap enough that people can afford it. If you look at what computing has done and how much more computing we get per thousand bucks than we were getting 30, 40 years ago, robots haven't dropped in price. They haven't gone up. They've beaten inflation, pretty much steady. But for the most part, robots are $30,000-$100,000, depending on the robot, to get it into consumer's hands. And the only company that has demonstrated they know how to get things into consumer's hands cheaply enough was iRobot when they came out with the Roomba vacuum. And that's an incredibly simple machine compared to what we're trying to do.

Mitch Weiss

And a high end iRobot, a high end Roomba is $900. There is competition coming from overseas for that, but those are not things that will work in fully unstructured random outdoor environments. So getting things cheap enough and reliable enough and easy to use enough is very hard. I think we're close. I think we have the right approach. Talk to people here and people I meet within the industry that what we're doing at PFF is probably the best, cheapest robot platform they'll ever see in terms of how we figured out how to make it consumer product priced, but it's going to take a lot of work to get that out there and to give it enough usability that people feel they're getting their money's worth.

Jeffrey Schnapp

You already mentioned the example of the Roomba, which is probably the most successful consumer robot to date. A lot of the consumer market robots have followed the path of kind of humanoid robots. They haven't been particularly successful in that regard, but I'm wondering, in the case of gita, our design team was very committed to not following a kind of humanoid model of designing the robot itself, we wanted to create a new category of vehicle that moved away from the kinds of expectations that a human form robot, a robot that talks, a robot that has eyes that it bats at you and so forth. I'm wondering, what do you think about that strong tendency to think of consumer robotics in relation to humanoid type objects?

Mitch Weiss

Well, I've designed my share of androids. I may be one, you never know. But why? The magic question is why? What's the utility of the thing? I was especially unhappy with the amount of work that goes on and has gone on in the social robotics space. I understand that there are plenty of people out there. I think it's super important and it's the right thing to do. I just don't agree with it. I don't see any value in it from a consumer product perspective. It's valuable as a research thing because at some point, we're going to share the world with these things. It ain't going to happen in my lifetime and we'll want to understand how people interact with machines better. But as a product to bring to market, to have in your house, it offers very little utility. And if you're going to give it to your kids, I think it's a lot less play value than a Lego or something where they can do something new with it every day.

Mitch Weiss

Those machines are always going to have limited responses and limited utility. One of the things that attracted me to PFF is what does gita do? Well, it carries heavy stuff, man. That's useful to me. It's got lots of utility. Now if we can get some personality and character out of its behavior and its design, all the better because then I'll be happy to have it carry my stuff. When given the choice between buying a four-wheel vehicle driven by motorcycle chains, an old Trabant or a VW or having a choice of buying a Ferrari, you're going to buy the Ferrari if you can afford it because you like it better. So there's no reason that we can't make it something you like and something that's useful. So gita carries two big bags of stuff and it's not much bigger than the two big bags of stuff, and we figured out how to do that, and that was really hard to do.

Mitch Weiss

And then giving it the smarts to do it smoothly and nicely where you don't have to talk to it or gesture at it or use an app to make it work, that was hard, and it's still hard. We're still making it better.

Jeffrey Schnapp

Yeah. I think what you're pointing to, which really is the core enterprise here, is creating a kind of human robot interaction that's highly intuitive, a kind of natural extension of existing behaviors that doesn't require a whole series of learned behaviors or adaptive behaviors on the part of the operator. And that seems like a good point in which to move towards the conclusion of our conversation today. And I'm just wondering what advice you would give to fellow roboticists who are tackling these kinds of problems today. What are the things that you've learned, the bitter as well as the sweet lessons, so to speak, that will shape the role of robots as we move into the next couple of decades?

Mitch Weiss

So the robot space, robot technology, I think, as an uber-nerd, is extremely cool. I've been doing it for a long time, like doing it, have lots of patents doing it, invented a lot of weird stuff, but it's always followed one simple question, why? What for? What's it useful for? I've never invented anything or built anything or tried to push on my customer base or anywhere else something that doesn't offer them true value and utility. So again, kind of begs to my social robots angst, if you're doing something and you're solving a problem and it's already solved, but you're doing it in an interesting new way, don't, because no one needs it. Do something that isn't being solved and do it in an interesting new way.

Mitch Weiss

So when you look at the consumer robots market, from the first one Nolan Bushnell came out with that had an eight-track tape in it, to today, what we're trying to do with gita is help you carry stuff around. It's got a function, there isn't a solved model for that, and we're doing it in an interesting way. So we get to leverage all the robotics technology that's out there today and satisfy our robotic nerd points. But if you're going to build another six degree of freedom manipulator to pick things out of a bin, you better be doing it a lot differently or a lot faster or a lot better than the other guy because there's plenty other guys doing it. Just because it's interesting doesn't make it important to the consumers of the technology.

Jeffrey Schnapp

Those are really some terrific thoughts on this topic. Thanks so much for joining us today, Mitch. Really enjoyed the conversation and look forward to continuing to develop gita and the successors here at PFF with you.

Thank you for listening to the Piaggio Fast Forward podcast and come back soon for further lively conversations about walking, light mobility, robots and the design of neighborhoods, cities, and towns. The PFF podcast is hosted by Jeffrey Schnapp, sound engineering by Robert Allen, narration by Ryan Harms, produced by Elizabeth Murphy, web designed by Jerry Ding. Intro music is Funkorama by Kevin MacLeod. End music is Your Call by Kevin MacLeod. Special thanks to Tory Leeming. To learn more about PFF and gita, please visit piaggiofastforward.com.