When Drones and Robots Knock on Your Door

Episode 4 Transcript

Jeanine Herbst
The next time you order something from your favorite store, pharmacy, or local restaurant, it could be a drone or a robot showing up at your door. That's because artificial intelligence is transforming the way companies deliver their packages to you. But is that a good thing?
 
Hi, I'm Jeanine Herbst, and you're listening to Where What If Becomes What's Next from Carnegie Mellon University, where we explore how artificial intelligence impacts industries, the government, society, and you.

On this episode, we'll look at the ups and downs of using AI, robots, and drones to get stuff delivered to you. Most of us are familiar with the trucks that deliver a sweater from our favorite retailer, groceries, furniture, and more. That method's pretty standard. Major retailers have central hubs and either truck or fly products to different parts of the country, where they're processed in a warehouse near their destination point. And that's not changing. What is changing is called the last mile. That means getting a product that last distance to the customer.

Challenges and Innovations in Last Mile Logistics

And that's the most expensive and time consuming part of the supply chain process, one that some estimates say could cost more than 50% of total shipping costs. Now, some companies are using AI to go that last mile and save money and possibly also save the climate.

Joining me today is Matthew Johnson-Roberson. He's the director of the Robotics Institute and a professor in the School of Computer Science at Carnegie Mellon University. He's also the co-founder of Refraction AI, an AI-powered last mile delivery company.

So Matthew, tell me about Last Mile Logistics.

Matthew Johnson-Roberson
So last mile logistics is probably one of these areas that we're seeing a lot of focus on automation because of the challenges associated with the costs and some of the externalities. So I think in the future, it's sort of our hope that we're going to see an increase in the number of robotic solutions in the space that kind of help to alleviate these last mile challenges. Because again, there's a lot of benefit to getting things through a distribution center.

But they have to get to people's homes or to the businesses or to other places in cities beyond where trucks can take them.

Jeanine Herbst
What does AI bring to the table that previous ways of distribution of home delivery didn't?  

Matthew Johnson-Roberson
Yes, I think the two big things that I think robotic delivery really enables is one, that it tries to reduce the cost. And that's probably one of the biggest drivers. But the second is that we think there's a lot of value from a sustainability perspective.

If you have a human doing last mile logistics, you need to build a big metal box around them to keep them safe, obviously. And as a consequence of that, the weight of the vehicles that you need to transport these things around in goes up significantly. And as a consequence of that, energy usage. And so as we think about the long-term sustainability of doing last mile logistics, one of the more important aspects of that is thinking about ways that will reduce the environmental footprint. And we've seen the really clear trend. People are getting more things delivered, not less.  And so if we're going to find a model that's going to work in the long term, we need to think about ways of making that, again, more environmentally sustainable.

Jeanine Herbst
So you're saying we'll be pulling the trucks off the road, so therefore the gas emissions and everything, the wear and tear on the streets and the bridges and such.

Matthew Johnson-Roberson
Yeah, yeah. So using robots is much smaller, lighter, and then you need to use less energy and again, less wear and tear on the streets. There's also challenges around congestion, right? So large trucks have a lot of challenges that they place on cities when it comes to congesting central business districts and things like that. And so all of these things are sort of pointed in the same direction. We're trying to think of other ways of doing this. And it's not just robotics, obviously bike messengers and other things are in the mix. Electrification of vehicle fleets is a big part of this as well. And certainly I would never argue that robotics is the only solution, but I think it's going to be additive to those other ways of doing this.

Refraction AI's Approach and Customer Feedback

Jeanine Herbst
So your company Refraction AI, which is available in three cities, uses a three wheel delivery vehicle, the Rev One robot. How does that all work?

Matthew Johnson-Roberson
So we build ground vehicles and they're tricycles, but they don't go on the sidewalk. They work in the road. So they're about the size of a bike, about the size of a person on a bike, way about the same. And they travel where a person on a bike would travel. So you can think of them as a robotic bike messenger is how I like to think about them.

Jeanine Herbst
And the robots deliver food?

Matthew Johnson-Roberson
Yeah, so this is coming from restaurants or grocery stores. So the two big things we're interested in are our restaurant and grocery. And those are both because you need to be home typically when those things show up. So your milk doesn't spoil. And that works really well with where the technology is. So people have tried a bunch of things and figuring out how they could build robots to climb up your stairs and get to your front door. But a lot of that technology really isn't ready for prime time yet.

And then if you think about the challenges around apartment buildings or big buildings that have a number of units in them, obviously the robot technology has not reached the point where it can go in there, press the elevator button and get up to your apartment door. And so again, sort of focusing on where the technology can meet the need. And there's sort of an immense amount of need around bringing food to people's homes and having them come out and grab it. It actually works almost the same as how your delivery process with Uber Eats or DoorDash works right now.

You know, you get a text message telling you that there's a delivery driver outside, and then you come grab your food and go about your day. So, you know, the compartment's locked and so you're given a code to unlock it and that's the mechanism we use to secure the food.

Jeanine Herbst
And the customers, what are they telling you about this?

Matthew Johnson-Roberson
So we've seen pretty positive feedback in terms of our customer experience, mostly because people like interacting with robots. They think it's cool. And so we get a lot of people that come out and bring their kids out and take videos and are pretty excited.

But then the second thing is also there's sort of another element to this, which is around just the sort of experience of getting your food out of a robot. Yes, certainly people don't like coming out of their houses, but then the flip side of that is we found that people feel pretty safe and comfortable with the robot and it's a more repeatable experience. They know what they're going to get. And then we can also sort of tighten up the delivery windows, right? Because we are sort of have complete control over what the robot does and where it goes and X and Y and Z.

And so one of the things that we also hear that's really positive is that people are pretty frustrated when it says your food's going to come in 20 minutes and then it comes in an hour or it comes in five minutes and you're not ready and you're in the shower and all of that. So the ability to sort of more accurately forecast and bring the food at the same time are the things that are, I guess, on the other side of the annoyance of having to come out of your home.

Future Outlook for Last Mile Delivery

Jeanine Herbst
So what does this all mean for jobs? And I'm thinking of the threat to gig workers, you know, Uber and Instacart drivers, for example. But can AI deliveries give a boost to local economies?

Matthew Johnson-Roberson
Yeah, so when we think about this, certainly in aggregate, the high level goal is certainly to increase the number of jobs, not to decrease them. Now, to your point, is there going to be some local displacement if this really scales up? Yeah, possibly. But ultimately, two things I would say. One is that our goal is to build out jobs that are sustainable, that have dignity, have a sense of purpose. 

And if you talk to gig delivery drivers, many of those things are missing. And then the economics around that, particularly on how much money they are actually able to take home, there's a lot of challenges associated with the current model. And so it is not as if robotic deliveries are placing good paying jobs with health care and a sort of lifelong career progression. Many drivers really struggle to make the economics work. And in fact, as a consequence of that, there's a great deal of churn. So drivers enter these platforms and then leave them because they can't make ends meet doing those jobs. And so we're really focused on trying to think about, you know, what does last mile logistics look like in the long term in the economy and how would we continue to advance that? And I think robotics plays a big piece. So what kind of new jobs will your company create?

My specific work is that we have people that provide on street support to the robot. So that's everything from maintenance to helping vehicles that get stuck to, you know, helping customers. But those are people that are able to sort of do a bunch of things, many of which require a higher level of skill. And so we have training programs and other things associated with that. And so they learn how to build and repair robots as opposed to just drive a car to drop food off. And so our hope is that not only does that sort of open up opportunities with us, but that there are a number of sort of robot technician jobs that are going to be coming online in the next decade. And so that these people have now sort of much greater opportunity, more greater mobility, and hopefully in the long term, greater economic opportunities.

The second is that there is sort of this remote oversight. And so we're able to kind of watch over the vehicles. And that's an air-conditioned job where you sit in an office as opposed to jumping in and out of your car.

Jeanine Herbst
You're listening to Where What If Becomes What's Next from Carnegie Mellon University. When we come back, we'll talk more about training for the new AI delivery jobs, the use of drones, and safety.

Steve Wray
As we enter the election campaigning season, there is the risk that generative AI will create deceptive and misleading content in real time and at scale and influence elections around the country. How can voters protect themselves and democracy itself from this misuse of AI technology? Hello, I'm Steve Ray at the Block Center for Technology and Society at Carnegie Mellon University, and we have put together a Responsible Voters Guide to Generative AI in Political Campaigning.

These guidelines will help you to identify deceptive and misleading content and give you action steps to protect the democratic process. You can download this free guide from the show notes to this podcast episode.

Jeanine Herbst
Welcome back. We're talking about AI and last mile deliveries with Matthew Johnson-Roberson. So you were talking about those Robotic trainings. Would that be at the local community college? Would that be from a company? Is that maybe even a space that, you know, somebody is going to start offering that training and, you know, start a company that way? What would that be?

Matthew Johnson-Roberson
Yeah, I think all across the board. So we're doing that training in-house at Refraction, but, you know, we're seeing lots of efforts here and actually at Carnegie Mellon, we've been thinking about this. And so there are trading offerings, particularly around robotic manufacturing that we're trying to scale up in partnership with community colleges here in the Southwestern Pennsylvania region. And so as we think robots are going to be in many more industries, not just last mile logistics, we're going to need people that are trained to both operate and maintain that. And we think that that again is a sort of a real growth opportunity and perhaps an area where we're going to see a bunch of retraining, a bunch of work with community colleges, a bunch of work with technical colleges to again sort of expand that workforce because we know that we have a real mismatch in the amount of labor that's available to do those kind of jobs, what we're perceiving to be a greatly increased demand in the next 10 years.

Jeanine Herbst
What's holding everybody back? What's the big road bump, speed bump right now?

Matthew Johnson-Roberson
So I think a lot of this is figuring out some of the technical challenges. Other challenges are finding a business model that's scalable and sustainable, right? So thinking about the economics of this and how would you put many, many robots on the road?

Jeanine Herbst
What are the technical challenges?

Matthew Johnson-Roberson
Making robots drive themselves is hard. So as we've seen in self-driving cars and a lot of other domains, right, you know, these challenges, everything from the sensors being better, more reliable, more robust, our detection algorithms and planning algorithms being you know, more reliable and robust, being able to handle edge cases. So thinking about things you've never seen before and making sure that you can handle all of those, ensuring safety, increasing speed. But, you know, as with any new technology, we're optimistic that if we keep pushing, we're going to get there.

Jeanine Herbst
So is this new technology safe?

Matthew Johnson-Roberson
Our first and highest priority is always safety, right? Getting your burrito is great, but, you know, most important is that we make sure that everybody gets there safe. So we've done a couple of things that really speak to that. Part of our fundamental philosophy in the robots, at least, that we built, my company Refraction has been thinking about how do we make them safe by design so they are much smaller, they're bike size, and they're much lighter, so they're not 4 ,000 or 5,000 pounds. And as a consequence of that, the risk profile of them is very different than, say, an autonomous car. If you get by an autonomous car, it's probably a pretty bad day. But if you're getting hit by a thing that is bike size that's going 10 miles an hour, it's probably not great, but also...

Certainly the risk for serious injury and fatalities is much, much lower. So that's thing one. And then thing two is building in the autonomy, the automatic emergency braking, all of the systems that you need to deploy this safely. And so we're really focused on making sure that we can respond really, really quickly. And then in scenarios where the vehicle doesn't know what's going on, we have the ability to jump in remotely with a remote operator who can take over. And so those remote operators keep an eye over the vehicles. And really the last stop for safety. And if they see something unsafe happening, they can jam a button and bring the vehicle to a halt. They can take the vehicle to the side of the road. They can do all the things necessary to place it in a safe configuration.

Jeanine Herbst
Well, what about drones for the last mile? I mean, they probably won't work in urban areas, but for rural communities, could play a big role.

Matthew Johnson-Roberson
I think certainly in rural regions and in other sort of domains like that, we're going to see a lot of benefit from drones. I think for urban delivery, there's a number of challenges that drones face.

Probably safety being the biggest, but also airspace and regulatory. But if you think about really rural areas, and particularly in areas where you need sort of mission critical things, so medicines, organs, blood, all those kind of things, I think drones are going to play a big role. The big challenge drones face are sort of weight limitations, right? So they can't carry very much. And then obviously takeoff and landing, and then if they fall out of the sky, that's bad. So those are all the sort of issues.

But again, I think in rural delivery, drones are going to play a big role. For rural delivery, you need takeoff and landing points or a way of dropping the goods. And so people have done everything from parachutes to cables that lower things to a second drone that drops down. But that sort of last 10 feet challenge is one of the bigger issues that drones have been facing in their sort of rollout and scale up. But you need a centralized distribution point. And if you're doing multi-rotor drones, then the ranges are actually more limited. You have 20, 30 minute flight time and so that presents its own challenges. If you're doing fixed wing, so more like an airplane, then you need a pretty significant runway to take off and to land. Those are the challenges that drones are trying to overcome.

Jeanine Herbst
Where do you think AI and the last mile delivery is heading, both as a professor and a business owner?

Matthew Johnson-Roberson
If you see the impact that artificial intelligence has had on other fields like language, you know, we think that the next five or 10 years in robotics could be incredibly transformative. And as a consequence of that, you know, we're trying to train the next generation of roboticists. And we think this is an incredibly important time. And it's a really exciting time for that.

From the company perspective, I'm really excited to continue to try to commercialize the stuff that we do in the research lab, just because it is probably the best and fastest means to get it to touch people's lives on a daily basis. And so my dream is always to look out your window and see a robot out there doing something useful. And we think that commercialization is one of the ways that we'll see that happen.

Jeanine Herbst
Thanks to my guest, Matthew Johnson-Roberson. And thanks to you for listening please check the show notes where you'll find links to additional resources for this episode. And to learn more or contact us with a story idea or comments, please visit ai.cmu.edu/ podcast.

If you're enjoying this show, please rate and review us on your favorite podcast site and share with your friends and be sure to subscribe so you'll never miss an episode. I'm Jeanine Herbst and you've been listening to where what if becomes what's next from Carnegie Mellon University where we're exploring what AI means to industries, government, society, and you.

Stay tuned and stay curious.