In automated warehouses, robots’ reach exceeds their grasp

Amazon, along with other online retailers, has seen a massive increase in demand in this time of coronavirus. The company recently hired 175,000 additional human workers, but its warehouses are highly automated, with robots moving giant towers of all the items we’re ordering around the warehouse floor.

But Amazon has also faced accusations about its working conditions being unsafe, especially during this pandemic.

I spoke with Ken Goldberg, a professor of industrial engineering at the University of California, Berkeley, and chief scientist at the warehouse operations startup Ambidextrous. The pandemic has increased demand to automate warehouses, but the technology isn’t quite there yet to improve conditions for, much less replace, human workers. The following is an edited transcript of our conversation.

Ken Goldberg: Robots are still quite klutzy. They have improved and we’re getting better, but the ability to pick up any object that you put in front of a robot, we’re still not there yet. Humans are very, very good at this. Even a small child —you put something down, any shape or size, and [they] can pick it up. Robots will often still today drop objects, and the reasons are really a bit subtle.

Jack Stewart: What are those reasons? We’ve been promised that robots would come in and take over all our house chores and automate our factories since the ’50s or ’60s, and it just hasn’t happened.

Goldberg: I know. I’m sorry about that. We haven’t gotten our jetpacks either. As I studied this field, there’s a fundamental issue [of] uncertainty in perception. Understanding where things are in space and uncertainty in control. Even if we knew where things are in space, getting the robots’ grippers to make the correct contacts in space [is difficult]. The third [obstacle] is uncertainty in physics. There’s uncertainty in the friction and the mass and the weight distribution of objects. All those things combine, so that there could be a very small error, even a millimeter error at the fingertip, and that can make the difference between picking an object up successfully and dropping it.

Stewart: Has this increasing push to automating warehouses because of COVID-19 helped speed any of this development up?

Goldberg: Yes, it’s accelerated the interest from industry and what has been already going on from the academic side. But one other factor that happened is that there’s a need to space workers out in the warehouses that exist today. There’s a real push right now: Can we start providing robots to insert them between workers in a warehouse to keep up with this increasing demand? I don’t believe that we’re going to replace all the workers in any foreseeable future. Humans have many good years left. We’re reaching a critical point where we can’t hire enough human workers, and we need robots to support them.

Stewart: Robots are always the sexy solution, right? But aren’t there other simple ways that we could use technology to help people stay reasonably well distanced as they’re working in these big warehouses?

Goldberg: Physical space is a challenge. Every inch is valuable in a warehouse, especially as you want the warehouses to be able to deliver on a shorter time frame. If you want people to be physically separated, then you do need some kind of automation there — conveyor belts, there’s rolling robots, etc., but the grasping is the bottleneck.

Stewart: What about just redesigning warehouses so that they’re made for robots as opposed to made for humans like they are now?

Goldberg: There is some thought about doing that, but one of the biggest problems is that may take quite a long time because there’s so many products out there. A typical warehouse may have 500 million different objects, and every order is different. Someone may order a mustache trimmer and a chew toy. We have to get those two products, they have to be picked up, and the shapes are vastly different. So, being able to solve [the] problem [of] universal grasping, being able to pick up an arbitrary object, that’s the grand challenge right now.

Stewart: There are very serious hopes around the world that a vaccine will be introduced for COVID-19, possibly within a year or so. Is anything really going to change that quickly in terms of automation, that it can make a difference?

Goldberg: I do believe that the adoption of e-commerce is going to be a permanent change. In other words, many people now are just accustomed to new patterns of shopping. They’re ordering many things that they wouldn’t have ordered three months ago. That’s going to continue, even if there’s a vaccine and we get back to normal. People are now comfortable, and so that volume is going to continue to grow. That’s where the bottleneck is. That’s where the challenge is, how do we actually get those orders delivered in a timely fashion to the customers out there?

Related links: More insight from Jack Stewart

Marketplace Tech’s Stephanie Hughes visited an Amazon warehouse in Edison, New Jersey, last summer. It was a fascinating tour of one of the online retail giant’s facilities, which it does open to the public for tours. Or did, before COVID-19 put a pause on that. We’ve got video too, so you can see these tall, yellow, robotic towers scuttling around. As Stephanie reported at the time, Amazon wants you to know that its warehouses are fun enough for the Girl Scouts to visit.

The tours are, of course, part of Amazon’s public relations effort. For all the use of automation, human workers have long complained that conditions are grueling. “I’m not a robot” is the way a Guardian story from this February headlined it.

February was in the before-times, without the added stress of the coronavirus, which both drove demand for online shopping through the roof and increased the risks for warehouse workers as they tried to fulfill those orders for us. There’s a lawsuit being pursued by Amazon workers who claim they got the virus at an Amazon warehouse and brought it home with them.

This feels like a situation where tech should be able to keep humans safe, but robots might not be there yet. Amazon is using image recognition software called Distance Assistant to monitor employees as they move around. Green circles are drawn around people on a screen to indicate a safe distance; red circles if they get within 6 feet of one another.