It was cold in the parking garage. We had been sitting there, off and on, for almost 9 hours a day, both Saturday and Sunday. Watching the robot tirelessly patrol the facility. It can be pretty boring, watching a robot roll from one end of the garage to the other, scanning for problems, relaying video. But we are testing the robot, so we need to pay attention, stay alert. It gives one time to think.
What strikes me, over and over again now, is the huge difference between a laboratory project and a product. When I was finishing my doctorate, I worked an a number of robotics projects, and we did hundreds of demonstrations on new robot technology. Each one carefully set up, and monitored by grad-students. Today, we are testing the Gamma Two Security Robot. It is designed to serve in empty warehouses, event centers, and other facilities, late at night when the place is (supposed) to be empty. It is intended to provide security for eight to ten hours a night, 365 days a year. Without having a protective grad-student technician trailing along behind to watch over the robot. And that is where the difference between a product and a project shows up.
The university setting is designed to support learning, and reward new contributions to knowledge. So every year, in hundreds of robotics labs around the country, students are doing brilliant research into new and better ways to make robots smarter and better. I spent six years doing exactly that, and did dozens of projects. Every few weeks, if you are looking for them, you can find press releases from this lab or that lab about their new robot that can fold laundry, or cook breakfast, or (perhaps the most popular) fetch a beer. The press release comes with impressive video showing the robot executing its task flawlessly. These videos became popular after the release of the Shakey video from Stanford. The robot planned and executed tasks and was under development from 1966 (yes – over 40 years ago) to the 1970′s.
Unfortunately, almost every one of those videos is more like a commercial than a documentary. The sad fact is that the 2-3 minutes you see on YouTube often requires several hours of painstaking set up: to get the robot into the exact location, the slices of bread aligned perfectly, and the precise orientation of the refrigerator door handle mapped into a CAD program. And then, even with all the high-tech laser measurements, it still might require five or ten, or even 20 or 30 ‘takes’ to get a clean video. Just like the perfume commercial, or the new car ad: “Cut!, set it up again, everybody else take 20!”
This is due, in part, to the academic system – almost nobody gets a PhD for making a system more reliable, or easier to deploy. So, as a researcher you are rewarded to do something new, and then something newer, and behind you lies a trail of projects that got good enough to make the video and write the paper. Now this is not universally true. There are labs like Carnegie Mellon where there was a museum guide robot named ‘Chips’ that has logged almost 500Km in over four years of giving tours.
That becomes the difference between a project and a product. The product has to work, every day, all day. Sure it will ‘call in sick’ with a bad sensor or a flat tire. But it needs to be dependable. Most projects work some of the time, good projects maybe 50% of the time, and great projects might work 99% of the time. But if your car only started 99% of the time, you would end up stranded about once a month. You are not going to buy that car again!
So here we are in the garage, watching the robot. We’ve logged over 120Km in the last year, but it is still not robust enough. We are working on our first five kilometer day. Five thousand meters of hands-free, no-problems patrol. And not in a lab, we are testing in a working parking garage. Cars come in and out, people walk by, every day someone has left a stack of boxes or a pallet somewhere unexpected. We don’t re-program the robot, the robot has to deal with these changes on its own; it has to show up every day for work, and it has to do its job well. Otherwise no-one will ‘hire’ the robot.
What drives us is simple. For fifty years they have been promising us robots. We see them in the movies, we hear about them on the news, but we are still asking “Where is my Robot?”
I do wish they would heat the garage.