With great expectations my coauthor, Tom Fitzgerald, and I sat down for a meeting with iRobot’s CEO. The time was early 2003 and, since just before Roomba’s launch the previous year, Tom and I had been working on a book titled Making it Go. Our publication would give readers a behind-the-scenes look at the conception and development of Roomba, the world’s first practical floor-cleaning robot. Earlier, when we’d told iRobot of our plans, the company responded positively. Today’s meeting was to review the details and, hopefully, secure the company’s official blessing.
AK Peters, the publisher of an earlier book of mine (Mobile Robots: Inspiration to Implementation) was supporting our effort. An inveterate robot enthusiast, I’d proposed the new book to them once it seemed likely that Roomba was actually going to make it to market. I had no idea then how successful Roomba would eventually become. My main interest was just to tell the story of the first behavior-based robot available to consumers. But because I had no experience with that genre of book (narrative nonfiction) the publisher introduced me to Tom, an accomplished author they knew, who had.
Tom and I hit it off and we immediately began working together on the project. It was great fun! I’d write long descriptions of events and Tom would turn that raw material into prose of a higher literary quality than I could manage. By the time of our meeting, Roomba’s sales were impressive. But the story we were developing was mostly focused on technology and the thinking behind the robot. We had completed three or four chapters developed along those lines.
Our meeting with the CEO that day began well. At first he spoke fondly about the story and the attention it might bring. But he kept on talking. And the longer he spoke the more he seemed to sour on the whole idea. By the end of the meeting he’d completely turned around. I’m always a little slow, but Tom saw right away that our project was doomed. Although iRobot might not forbid publication, they would make it difficult for us to proceed. (E.g., iRobot’s chief council told me, “You can’t say that Roomba is a success.” What!?) Tom walked away that day.
I now better understand why iRobot was wary of our book. Every company wants to be the only one telling the story of their own products. Third parties—especially third party insiders—can complicate the message the company is trying to tell. But at the time I was just baffled as to why the company would scuttle a favorable story and free publicity.
But, with Tom gone and the displeasure of management made clear, I too abandoned the project. And there things sat for almost 20 years. After leaving iRobot, I occasionally considered reviving the book. But, it always seemed that my new enterprises demanded all my time, leaving none for writing.
Eventually, my other work wound down, and I more or less retired. That meant I could finally give Roomba’s story the time it needed and deserved.
In the years since my earlier books, AK Peters had ceased operating and the publishing industry had transformed in ways not especially comfortable to authors. But I found some fortuitous help. A fellow author who had interviewed me for his book (Oliver Mitchell, A Startup Field Guide in the Age of Robotics and AI) introduced me to his publisher, CRC Press. CRC accepted my proposal for Dancing with Roomba and welcomed me into their merry band of wordsmiths.
Research
The beginning of Roomba dates to 1985-86 when Prof. Rodney Brooks formulated his ideas about behavior-based programming. That new paradigm for robot control greatly facilitated Roomba. My earliest involvement in robot floor-cleaning came at the MIT AI Lab’s Robot Talent Show in 1989. Roomba was developed at iRobot over the course of three years starting in 1999.
My memory of that period was imperfect—I’d forgotten many details from those days. Also, beyond my fading recollections, there were numerous Roomba happenings outside my knowledge. My focus was on technology, so I often missed key Roomba events that took place in the non-technical realms that I rarely visited.
Thus, I needed help. But from that necessity sprang the most fun part of writing Dancing. In August 2022, I sent a message to the other six members of Roomba’s core team telling them of my publishing aspirations. Everyone was instantly supportive! My enthusiastic teammates pitched in to patch holes in my memory and to relate stories—some of which I’d never heard or even suspected. They, and a few other former iRobot folks, let me interview them for the book. These encounters were always tremendous fun, filled with reminiscence and revelation. I’m very grateful for their time.
And one other bit of magic occurred. The Roomba team was shocked when, during our development, deep pocketed conglomerate Electrolux launched Trilobite, the world’s first home robot vacuum cleaner—beating us to the punch. Electrolux’s robot competed directly with Roomba and, although Roomba ultimately prevailed, the team sweated greatly over Trilobite at the time. The magic was that I managed to track down the leader of the Trilobite development, Per Ljunggren, and he agreed to help me! Per and another Electrolux engineer, Ulrik Danestad, were most gracious in sharing their robot vacuum experiences with me. Although I had no inkling before speaking with Per, Trilobite’s tale turns out to be as unexpected and fascinating as Roomba’s.
In the years before Roomba, I was well aware that there had been earlier attempts to build robot vacuum cleaners. But trying to quantify those efforts for the book was eye-opening. I discovered no fewer than two dozen serious projects. These weren’t only the work of basement tinkerers dreaming of robot riches. Most would-be inventors were, like Electrolux, well known corporations—folks serious enough to devote significant engineering time to the endeavor and to incur legal costs to secure patents. Starting around 1956 these attempts went on for nearly half a century, until Roomba launched in 2002. All those failures argue that our attempt to build Roomba was even more audacious than we realized.
Writing
There may be authors who can produce an article or a book in a single draft. Folks who have the clarity of thought to work continuously from beginning to end with minimal revision. That’s not me. My work always proceeds in fits and starts. I write a long piece. Then decide that it’s poorly arranged or doesn’t say what I intended. Next, I make an outline or compose a list on the white board or maybe start writing again on a different tack.
That was very much my process with Dancing with Roomba. I suffered through many false starts and much rejiggering. As the photo shows, one trick I tried was to make a card for each section I was writing for the book and then rearrange the sections to see what order worked best. Along with other information, the cards were color coded according to their technical versus narrative content.
This step came after about six months during which I just tried to write as much as I could each day. For the statistically minded: I averaged 615 words per day with a standard deviation of 446. (Of course I made a graph, but I’ll spare readers that detail.)
Another thing that helped was having an original Roomba available to jog my memory. Long ago my home held many Roombas and pieces of Roombas that I used for testing. But over the years they somehow all vanished. Fortunately, Roomba’s EE, Chris Casey, is a more accomplished curator of historical artifacts than I am. He loaned me a couple of original Roombas, providing me with the molded-in-plastic and soldered facts of our design.
Early in my writing, especially when I got stuck, I occasionally consulted one or more AIs. The AIs never flirted with writer’s block. In seconds they always came up with a coherent paragraph or two—even if they had to make something up. I didn’t use any of the AIs’ suggestions directly, but they were sometimes helpful in illuminating approaches different from the one I was stuck on. However, these consultations became rarer as I wrote more. The AIs always seemed to give me a sort of statistical average of what a competent writer would say, but never quite what I would say. (Of course, an appropriate statistical average from its training set is exactly what an AI is designed to deliver.)
Even before the manuscript was done to my satisfaction, I began sending out versions to friends and family for their review. In exactly the same way that home testing vastly improved Roomba, friendly reviews changed Dancing in big and important ways. For example, initially the story of the RCA Whirlpool Miracle Kitchen was buried deep in the book. But I think the story flows much better given its current spot.
In mid-April 2025 after much checking and rechecking I sent the manuscript to the publisher.
Marketing (Shudder)
Once upon a time, I believed that a great idea was all one needed. Build a better robot (or write a good book) and the world would beat a path to your door. Fortunately, those callow days have passed, and I now appreciate that an idea is just one component of a project’s success. Indeed, a good idea that no one knows about has no more practical value than a bad idea.
And these days spreading the word about an idea or anything else usually involves a website and social media. So, with great trepidation, I built a website (DancingwithRoomba.com). Happily, it turns out that modern tools—although initially intimidating—make building a website much simpler than it used to be. The result is also better looking than what I could once have achieved.
Left for the future will be wrestling with social media and absorbing the verdict of the marketplace.
Thanks for reading and please stay tuned.
