Once upon a time a two friends stood in front of an overly complex trash can at a WholeFoods. They found that deciding what to do with their garbage was far more complex than it needed to be. Which bin does the paper plate go into? And is this a compostable fork, a recyclable fork, or a fork destined for landfill? Rather than simply dump their trash and walk away, they came up with an idea. A short time later CleanRobotics and their first product, TrashBot, were born.

They first launched TrashBot in the Pittsburgh City Council building in 2017. Since then, CleanRobotics has been named one of the top Pittsburgh companies to watch in 2018, has advanced to the second round of the IBM AI Xprize, and been recently featured in GeekWire.

I had an opportunity to sit down with Tanner Cook and Grant Halleran from CleanRobotics to discuss the world of trash processing and the design of TrashBot. Throughout the interview we discuss how it is especially challenging for a robotics company to balance the demands of engineering and product iteration with sales and market penetration. We also discuss how dealing with trash is a complex, yet fascinating, societal and environmental problem.

How did the idea for CleanRobotics come to be?


The idea came from an experience that Charles and Vaish, our founders, had at a Whole Foods. They went to WholeFoods to have lunch, and when it came time to throw away their plates they found that they were staring at a huge trash bin with six holes for compost, recycling and trash. And there was this collage of stuff on a picture behind it that was supposed to help you decide which bin to throw your stuff into.They were looking at it and were terribly confused. Like there are the compostable forks, but then there are other types of forks that go to landfill, and then there are some utensils that are recyclable. It’s just really confusing. And they had a thought that throwing away your trash really shouldn’t be this hard.

And the more they researched and dug into it the more they realized there is a huge problem with the way we deal with waste in general. We are actually, really really bad at separating what is recyclable from what goes to landfills in the United States. This is for a variety of different reasons.

And when you look at the solutions people are proposing a lot of it is being pushed on the back end. These large scale materials recovery facilities. But none of them are actually making any money. They are usually in place because of legislation and regulations and for companies to save face or promote an environmental brand. But none of them are really making money.

Why is it so hard to solve the problem on the back-end like that?


The big issue is massive amounts of contamination. Contamination being, stuff gets sorted for recycling when it is not actually recyclable. Or mixing of things that shouldn’t be mixed. Like when someone dumps that half-full coffee cup onto a bin filled with recyclable paper – that ruins everything. The paper can’t be recycled anymore.

So we were looking at further up the chain. Like looking at that moment when the coffee cup goes into the trash and keeping it away from the recyclables. We wanted to maintain the quality of the recyclables.

To play devil’s advocate, why can’t we just work on humans being a little less impatient with sorting their trash? Why do we need trash-bot?


It’s actually not just impatience. It’s actually a lot of times a misunderstanding about what is recyclable or not. Like for the longest time I thought coffee cups were recyclable, but they aren’t.

And it just takes one person to throw out a full cup of coffee to ruin everything underneath. So with TrashBot you can throw in that coffee cup and it never touches the recyclables. When you open our recycling bin in TrashBot everything is nice, dry, recyclables.

Tell me more about how trash gets dealt with. Where are the big pain points?


Any big to medium facilities, like shopping malls, airports, convention centers, will have a trash sorting facility within it. They’ll employ people to actually pick through the trash that gets thrown away and sort it. So then when they take their trash to a processing facility, they will get more money for their recyclables.


So they have to pay for the infrastructure to do this sorting, and pay people to do that job. And it is a dirty, stinky job. No one really wants to do it. And they have a huge problem with turnover. They will pay people $30 an hour to do that job, but they still have 60% turnover. So it’s a very expensive cost.

With TrashBot, we hope that they won’t have to spend as much time and money dealing with their trash. And our ultimately goal is that in the future they won’t have to deal with it at all. That all their trash will be perfectly sorted and ready to go right from the moment it gets thrown in.

Did you always know that you were going to build a type of trash-can? Or did you consider other solutions at different touch-points?


Yes, we did. But the more research we did into how difficult those problems are and how many players are already in that market space, the more we realized that there was a big opportunity to attack the problem from the front. That is, solve the problem from when the trash first enters the trash can.

Also, with the state of where technology is, with regards to computer vision and robotics, we found that we were able to build a TrashBot at roughly the same cost as what those facilities are already paying for their trash cans.


I think we always sort of knew that we were going to build a robot of some sort. But that part of the business of where we were delivering value to those medium sized facilities evolved over time. Because, you know, going back to the Whole Foods story, at first we thought the value was that people wouldn’t have to sort their trash. But as Charles, our CEO, started doing more research he found that there was way more value in trash-bot for the facilities.

So you are in an interesting spot, because obviously you need to design TrashBot to be usable for the people who deposit their trash. But it is being purchased by the facilities managers, so its gotta be designed for them too.


Right. And there are all types of users. You’ve got your custodian. You’ve got your building operations managers. Then there are the folks who handle procurement. And then there are of course the people throwing away trash, which is, everybody.

Wow – I don’t think I ever realized that trash goes through so many people’s hands!

Tanner: Oh yea. Trash gets dealt with on so many different levels.

Grant: And so we’ve got features to TrashBot that are designed to benefit each party.

So taking it back a little, once you had the initial idea, who did you start talking to? How did you approach getting into the trash business?


So we had lots of conversations with facilities managers and looking at their issues and building out our value proposition. And one thing that came out from these conversations is that commercial trash-cans that these people have to purchase cost a ton of money. And they are just steel tubes.

So that’s where the opportunity was. Everyone needs trash cans. They are expensive already. And we felt that we could build a solution that fits within their existing budget, but that also is more beneficial and helps these facilities on their backend trash problems. And TrashBot makes their lives easier in other ways.

So talk to me about how you get TrashBot to work.

Tanner:  Well, we’ve done trash sorting ourselves actually.


We get trashed delivered from the facilities that we are going to, so then we can train TrashBot on the actual trash that will be coming from that facility.

So like for the Pittsburgh Airport we got a delivery of Starbucks trash, and it was nasty.

Tanner: So so bad.

Grant: Stinky, decomposing coffee grounds.

Oh man, that sounds smelly.

Tanner: It is. All trash has a smell.

Grant: Yea and we just put on the elbow-high gloves and dig in there.

So you’ve reached this point where TrashBot that helps with the trash and recycling problems these facilities have. What is next?


We’ve done some focus groups and studies and found that people are willing to engage a little bit more with a robotic trash can than just a normal one. We’ve found that people will spend somewhere between 7 to 8 seconds engaging with a robotic trash can than the 1 to 2 seconds they’ll spend with a normal one.


And this is really great for us currently because TrashBot isn’t as fast as we want it to be just yet. So the fact that people are willing to engage and wait a little bit means that TrashBot is better able to do its job.

So how has your user research changed how you’ve designed TrashBot?


We participated in PGH Labs and launched TrashBot in the city council building, and our experience there showed us that the human-interaction piece is something we really need to spend time on.

We’ve redesigned trashbot a couple of times. We used to have a front-and-back design but there were too many degrees of freedom mechanically and also if you weren’t quick getting your hand out of the way the door could come down on your hand – which is definitely not good! So the latest side-by-side design fixed a lot of those ergonomic problems.

But one thing we found is that people will put in their trash, the door closes, trashbot does it’s thing, and then the door opens again, but the person is still standing there. They are standing there and aren’t sure what to do next.

And when we asked people about that experience they said all kinds of things. Some people wanted to know where their trash went. Some people wanted something to say “thank-you” so they knew they were done. We even had one woman say she wanted kittens and rainbows when she recycled as sort of a reward.

A picture of all the different designs of TrashBot. Starting with the original wooding prototype, then a prototype hacked out of an outdoor grill, then the front/back design with recycling logo, and finally the current version with side-by-side trash deposit holes.
The evolution of TrashBot designs

So it sounds like you are starting to get into the front-end design and experience of trashbot. Is that what you think is the next on the horizon?


That and handling compost. We’ve started to have conversations with west-coast facilities and out their compost is a bigger concern. So we are looking into how to design trashbot to handle three different types: recycling, compost, and regular garbage.


And generally, people are a little bit better at identifying what is compostable. Like, if it is something you eat then you know it can be composted. So we are looking into how to take advantage of that while integrating it into our existing ability to sort recyclables from trash.

What do you think is the biggest sticky, challenge you all are facing as a company?


As a hardware company, it is hard to balance the engineering cost and wanting to do iterations on the product versus the need to do production and make sales. It is a startup and sales are absolutely key to success, but it isn’t software. Doing an iteration on a robot costs a lot of money and time. So I feel like we are really having to balance the need to continually improve the product while also maintaining that sales pipeline and penetrating different markets.


Also it is much harder to gather user-data on a robot versus if we were building a website or something.

So how do you approach getting user-interaction data?


We’ve done interviews, where we interview people right after they’ve used it. But that is a bit hard to do and limited. We’ve also done surveys.


We have found that it is pretty easy to just watch people though. It’s easy to sit somewhere and just watch people use it as they go by. Thats been a good way to gather some data. That’s how we found out that people were standing in front of TrashBot after they were done and waiting for something else to happen.


And actually when we get the deliveries of trash from facilities, that has been a great learning experience. Like when we got the delivery from the airport, we found that one person had recycled an entire stack of newspapers that were tied up. It was the paper from the day before that just must have never gone out. But then a few people threw full water bottles and coffee cups into the recycling and it soaked the entire stack of paper. So that paper could have been recycled, but then it got contaminated. So when I think of what TrashBot can do design wise I wonder if there could be an education component and prevent that from happening.

 I bet you guys see trash cans and recycling bins everywhere you go now.

Grant: Oh yea, we study them.

Tanner: We have a huge shared folder of nothing but pictures of trash cans. We are always studying trash can design.