Projects don’t follow straight lines. I’ve never participated in a development that proceeded logically along preplanned steps from beginning to end. Always there are false starts and dead ends. Roomba was no exception. The product eventually purchased by millions of customers differed in several significant ways from the robot Paul Sandin (my initial co-conspirator) and I conceived of and proposed to iRobot. Beyond other features described in Dancing with Roomba, here are three curious items that failed to make the cut.
Deployment Stick
One concept Paul and I obsessed over was convenience. The robot would be promoted as a time and labor saver and we were sure no one would buy it if it failed to fully delivered on that promise. We wanted Roomba to be every bit as easy to deploy as a broom or upright vacuum cleaner.
That’s a tall order. All the owner has to do to use a battery-powered upright vacuum is to grab it from the closet, apply it to the dirt, and then return it to the closet. A broom plus a dustpan with a long handle requires a similar effort. Thinking about convenience, we noted that there is no stooping over in either of these scenarios.
But Roomba would be quite squat. That unfortunately meant that to use Roomba an owner might have to bend over four times—first, to remove it from the closet, second to place it on the floor, third to retrieve it from the floor, and forth to replace it in the closet.
This led us to commit a large amount of time pursuing a method that would let users deploy Roomba without bending over. We actually came up with something we called the deployment stick. It was just that, a stick or pole. About as long as a broom handle, the stick would let the user maneuver Roomba while standing upright.
When an owner wanted to use Roomba, he or she would first grab the deployment stick, presumably left standing in the closet where Roomba was kept. A feature on the bottom of the stick would mate with a receptacle on the robot’s top. With the deployment stick attached to the robot, the user could position Roomba as required. Next, the user would disengage the deployment stick and then use it to press the button to turn on the robot.
The concept seemed straightforward and not too difficult to implement. But we wrested with an existential issue: was the benefit of not having to bend over truly worth the cost and hassle of managing the accessory? Customers would have to store and keep track of the stick, and the Roomba team would have to find a low-cost way to box up the (collapsable?) stick in the same package with the robot.
An important strike against our concept was that no one we described it to expressed much enthusiasm for the idea. Ultimately, we couldn’t convince ourselves that the deployment stick was worth the trouble.
Traffic Cones
We knew early on that there would be places users wouldn’t want Roomba to go. For example, maybe the robot should always avoid the play area strewn with kids’ toys or the messy patch around the pet’s food and water bowls. Alternately, users might want to keep Roomba confined to an especially dirty area so cleaning could be accomplished faster. How could we create robot exclusion/confinement zones? The simplest way would be a physical barrier of some sort. A physical barrier had the significant advantage that Roomba’s bumper could detect it, thus freeing us from the burden of developing any other mechanism.
The first thing we toyed with were draft stoppers—those (usually) cloth cylinders filled with sand or other material that prevent cold drafts from seeping under closed doors during the winter. We borrowed some draft stoppers from iRobot employees, built a few of our own, and conducted some tests. The results weren’t good. The barriers worked well enough on carpet but on tile or hardwood floors Roomba wouldn’t leave the barriers where we put them. More often than not each time the robot bumped the barrier it nudged it. Eventually, it would push it out of the way entirely. We might have engaged in additional engineering by say, adding sand to make the barrier heavier and harder to push but that would make the Roomba box heavier and more costly to ship.
A cute idea that occurred to us was to include a few miniature orange traffic cones with the robot. Made of high friction material so they wouldn’t be dislodged, the user would position the cones—ensuring that they were less than a robot diameter apart—to block off any place the robot shouldn’t go. Cones would be stenciled with the words “Robot at Work.”
For several months the small cones became our default confinement scheme. But ultimately, they too fell by the wayside. Although adorable, the cones seemed cumbersome to set up and remove and including enough to block off a wide stretch created the same weight and cost problems as the draft stopper idea.
In the end infrared-light-emitting virtual walls became our production confinement solution. But we were interested to note that another group came up with the same idea we had. A couple of years later, after Roomba became successful, The Sharper Image company commissioned their own robot vacuum cleaner, called the eVac. eVac was supplied accompanied by four little traffic cones to confine it.
Easy Empty
Roomba provides a great user convenience by collecting dirt from the floor. But once this is done, the dirt is contained within the robot—what’s next? Somehow, the user has to move the dirt from the robot to a trash can. Paul and I thought that step might be perceived as inconvenient or unpleasant.
We tried to devise a scheme to mitigate the nuisance. The idea that emerged involved a big “dump” button on the top of the robot. Once the robot had completed its duties, the user would pick up Roomba, position it above a trash can, and press the button. This would cause the dirt container to rotate downward emptying its contents into the trash. Release the button and the container would return to its operational position. No muss, no fuss!
Unfortunately, we couldn’t engineer a simple way to accomplish that function. Many mechanisms needed to be packed into Roomba’s tiny internal volume. A purely mechanical system that would dump the dirt seemed likely to interfere with the main cleaning brush and other components. Although we liked the concept our finite time and resources resulted in its being triaged away.
We ended up with a straightforward mechanism—a removable dust cup. Pressing a tab on the back released the cup from the robot. The user would then slide the cup out, dump its contents into the trash, and return the dust cup, clicking it in place. Not exciting, but workable.
Eventually, iRobot developed a high-end version of Roomba that included a powered dock. When Roomba came home to recharge, the dock would suck the dirt from the robot and deposit it in an external bag for easy disposal. Even very early in the development of the robot we realize that sort of device would offer the ultimate in convenience. But, of course, auto-emptying was a mechanism too far for the first version of Roomba.
