...Starting with first principles
America First Trust services
Education, analysis, and advice for uncertain times



[Tetra Vaal ©The Embassy Visual Effects at www.theembassyvfx.com ]

Entrepreneur by Bill Fox


Part 2


First posted Feb 2005

During his talk at the March 9-10th robot conference in Cambridge, Dr. Rodney Brooks of MIT expanded on his sense of computer industry-related déjà vu.

He pointed out that In the 1970’s computers were large, expensive systems, often locked away from the public. Robotic applications are now leaving secured factory floors and are fragmenting into such areas as consumer products, mine-mapping, entertainment, and surgery. We see growing popularity with hobbyists, evidenced by the Battlebot competitions, local events such as the Seattle Robothon, rapidly expanding international organizations sponsored by major corporations for high school students such as FIRST, and the ROBOlympics for enthusiasts of all ages. Robotics departments at universities are gaining more research money and are creating more spin-off companies, particularly around MIT, Stanford, and Carnegie Mellon.

Dan Kara, the President of Robotics Trends who hosted the March 2004 robo conference in Cambridge, MA and Oct 2004 RoboNexus in Santa Clara, CA outlined inflection point criteria in his presentation "Sizing and Seizing the Robotics Opportunity." They include the maturity of enabling technologies, the extent of quantitative research (academic and corporate), the existence of real social pain points that motivate robotic development, the amount of government-mandated service/support, the degree of coverage by media, the diffusion rate of innovations, and finally, the level of participation by leading vendors, the investment community, and by smaller entrepreneurial companies.

According to Mr. Kara, we are now seeing the early adopter phases of the service robot and personal robot areas.

Whereas fixed robots have functioned in highly structured environments with little autonomy, service robots are designed for environments with medium structure and medium autonomy. Personal robots are geared for loosely structured environments with high autonomy.

Major sample service robot sub-markets include mine detection, exploration, unmanned vehicles, healthcare, robotic surgery, industrial cleaning, search and rescue, and security.

Major personal robotics sample markets include education, entertainment, vacuum cleaners, personal assistance, elder assistance, lawn mowers, and toys.

According to Mr. Kara, each decade of robotic development has been fueled by the academic research that took place in the preceding decade.

According to a chart provided by the National Institute of Advanced Industrial. Science and Technology, Japan, the first boom involved assembly line robots in the 1970’s. Academic research in that decade, which involved manipulation, 2-D visual sensors, and SCARA robots led in turn to the 1980’s when robots became entrenched in industry with arc welding and visual sensor capabilities.

According to the Japanese Institute, academic research into intelligent humanoid systems in the 1990’s, particularly in virtual reality-based remote control, behavioral intelligence, miniature on-board computers, Kansei robots [or "technology of the emotion" robots -author], and multi and micro robots, will likely fuel a second robot boom in 2000’s. The Japanese Institute envisions advances of robots into new fields that integrate robotic technology with information technology, create humanoid platforms, and provide intelligent entertainment.


According to Honda, "ASIMO rings the opening bell at the New York Stock Exchange (NYSE) on February 14, 2002, in honor of the 25th anniversary of Hondaís listing on the NYSE."
"ASIMO challenges a group of young students to a balancing contest during the `Say Hello to ASIMO' educational tour, visiting science museums and institutions."
"Curious fans join ASIMO on stage. The show is designed to educate children about robotics and to encourage children to study robotics and science. "
"ASIMO is the only robot of its kind to climb and descend stairs. For each step, ASIMO uses all of its sensors and computing power to counterbalance, adjusting all along the way."

Source: Honda Asimo

In “A Culture of Robotics “ Mr. Kara reported that according to the 2002 World Factbook, Japan possessed 410,000 of the world’s 720,000 “working robots” at that time. The Japanese appear to be working out the bugs on robotic concepts in their home market before blitzing the West. Mr Kara wrote: “One glaring difference between the West and Japan is the level of resources and support large corporate entities are dedicating to personal robotics. In Japan, Honda, Sony, Fujitsu, Sanyo, Hitachi, NEC, Mitsubishi, Epson and others (Can Toyota be far behind?), a veritable who’s who of corporate giants in the consumer electronics and automotive markets, are developing personal robots, humanoid and not, for use in the home. Honda alone has devoted 10 years, and spent a staggering $100M, developing its ASIMO humanoid robot. [Author's note: Honda informed me that $100 million is too high a figure; nevertheless the investment has been substantial, and in my opinion admirable compared to most American counterparts, who I regard as relatively short-sighted] For Honda, the plan is to develop and sell affordable and practicable humanoid robots for use in the home within 10 years..."

Image courtesy of
Sony Corporation

Before you dismiss Honda’s ASIMO project or any other Japanese robotic iniatives as "Carney" amusement items, take a look first at Sony’s AIBO (Japanese for “Pal” or English for Artificial Intelligence Bot) which was launched in May 1999.

The full-featured ERS-7AIBO model retailed for $1,799. Sony came out with a cheaper LM series (ERS-31L) for $599 in May 2002, and has sold out its limited edition runs. However, focusing on price probably misses the bigger picture. Sony will probably never recoup its R&D money on any of its current model robots, and probably does not care. In all likelihood, these robots are merely test market check points along the way of a very long term and comprehensive strategic plan that should have more Americans paying notice. The end game will come when Sony can flood the world at a profit with robots below $200 apiece for the consumer novelty-gadget market, and the closest competitors outside of Japan are years behind. Something similar has happened before with the Sony Walkman and other electronic consumer items.

Sony deliberately designed the appearance of the robot to be ambiguous, so that people could see it as a cat, dog, tiger cub, or any combination thereof. Somehow the “it is whatever you want it to be” approach did not quite wash with the public. So Sony has now made it official. It is a “dog.” Another item on the market test checklist.

The more important point is that its behavioral architecture makes it appear to grow through stages such as a newborn, baby, adolescent, young adult, and adult. According to Gareth Branwyn, “Each stage [offers] more interactivity, more sophisticated behaviors, and several different personality tendencies: sheltered, nice guy/gentle, adventurous, or selfish. Again, these are only tendencies, not fixed behavior sets. Each AIBO matures slightly differently, depending on how it was `raised’…It is amazing how little of this type of divergent behavioral programming is required before human beings start thinking they see intelligence…The current top of the line model can respond to 75 voice commands, has wireless capabilities, a camera that can take digital images on command, and 21 colored LEDs [Light Emitting Diodes] that it uses to communicate its `feelings’”.

Image courtesy of Sony Corporation

And now QRIO (pronounced “curio.”). 23 inches tall, and as yet un-priced, this builds on the technology created for the AIBO. It debuted in Sept 2003. According to Sony, “Seven microphones in its head identify a person’s voice and the direction from which it is speaking, and even pick out the words it sa ys…It knows tens of thousands of words already, but can also learn new ones. This technology is widely employed in machines other than robots (e.g. in car navigation systems).” (Branwyn comments: “If you’ve ever tried one of the computerized therapist programs that have been around since the dawn of the PC, or an online `chatterbot,’ you know how this goes. There’s enough conversational give and take that you can almost imagine you’re talking to a real person.”) Sony continues, “QRIO can have an entertaining conversation with you. It analyzes the words you speak using its voice recognition technology, and responds with its own words. It will ask what sort of things you like and remember them, getting to know you better all the time…”

The robot also sings, dances, and can kick a little ball (please note videos). It is the first humanoid .robot that can run. According to Sony, “QRIO has its own emotions –and expresses them in a variety of ways, such as through its movements, actions, sounds or colors. Sometimes, since it has its own emotions, it might not do something you ask it to do. It’s all part of the mystery of QRIO.”

Mystery indeed. Later in this paper I will discuss how Americans have also delved into various mysteries. One example is Kismet, the MIT robot project that engaged in emotional interaction with people.


The fist successful interactive “smart toy” in the mass consumer market was Furby, released in winter 1998. According to Gareth Branwyn, “It had simple sensors under its fur, and a motor that animated its eyes and mouth. It didn’t learn, but it gave the impression of learning by using a clock timer by using a clock timer in its micro controller brain to create seemingly random behaviors, unique personality quirks (every Furby was different) and speaking skills that appeared to improve over time. Legions of Furby owners, both children and adults, were awestruck by the apparent sophistication of the critter, ascribing to all sorts of lifelike behaviors it did not possess.”

The robot was invented by an American named David Hampton. He sold the toy to Tiger Electronics in 1998, which in turn was acquired by Hasbro.


In terms of innovative product development and financial survivability, one of America’s closest corporate counterparts to Sony’s mobile robot department is a privately-owned company based in Burlington, MA called iRobot. The company was co-founded in 1991 by Dr. Rodney Brooks and two of his former MIT graduate students, Colin Angle (now CEO) and Helen Greiner (Chairman). IRobot was profitable and made $55 million in revenues in 2003, half of it from military sales. Venture capitalists liked the story enough to invest $28 million in that year, and in 2004 the Army awarded iRobot a $32 million contract. IRobot and its product-market experience provide good insights into the unfolding nature of the mobile robotics industry.


According to a now removed iRobot/Hasbro web page:

"...IT was the first emotionally-expressive, emotionally-responsive robot, built in 1995. `IT' featured a very mechanical-looking face, and robot arms. `IT' loved shaking hands with people, and would smile and ham it up for anyone taking his picture. `IT,' however, would get frightened if you got too close to him and would actively try to avoid having light shined in his eyes....[BIT] was created the next year, and featured the first use of simulated facial muscles to make him smile, frown, and cry. `BIT' was designed using the Behavior Language to be just like a real baby, and although he could not sense people touching him, he could tell if he was upside-down (which he didn’t like), or right-side up (which he did like)....When Hasbro and iRobot began collaborating in 1998, the MY REAL BABY doll was the first toy the two companies decided to develop together – quite literally, the `first-born.'"

Colin Angle signed a three year exclusive with Hasbro in 1998, in which iRobot received $1 million to turn Bit into My Real Baby. Once Hasbro broke even on its investment, iRobot was in for a 5% royalty. The royalty business model was probably appropriate given the level of risk of the doll industry. According to Dr. Brooks it was tough designing components cheap enough in an industry where the cost of components are typically 6% of the sales price. (Flesh and Machines, p. 110). It was also hard to differentiate to the public all the robotic features on the doll when conventional wisdom in the industry had it that a thirty second TV ad normally allowed only enough time for one new idea. Lastly, the doll industry is not only subject to fickle fads, but remains very competitive.

The introduction of My Real Baby in bulk around $99 in retail stores in mid November 2000 encountered numerous merchandising problems that had little to do with robots. According to one source, kids did not see many commercials before it came out. Hasbro ran into production problems and missed the annual purchase planning cycle update of various retail stores. My Real Baby's newsworthiness got displaced by the Bush-Gore Presidential election crisis. In the toy business there is no such thing as a sleeper product, so slow sales caused many stores to make fairly immediate mark-downs as low as $59.00. Hasbro stopped production at around 100,000 dolls.

In 2001 “smart toys” comprised only 2% of the $25 billion toy market. Competition became a major issue early on, as Hasbro raced to beat Mattel's introduction of Miracle Moves Baby and MGA's My Dream Baby. My Real Baby is no longer in production, and currently trades for around $96-$99 in after markets such as eBay. It is still considered the smartest doll by Children’s Software Revue magazine.


The Roomba

And then came the Roomba: This concept reflected the philosophy underlying the article"Fast, Cheap, and Out of Control: A Robot Invasion of the Solar System” that Dr. Rodney Brooks co-authored in 1989. The idea was that rather than send a 1,000 kilogram payload to Mars at a cost to taxpayers of about $12 billion, NASA would be better off sending one hundred faster, cheaper autonomous little robots (using subsumptive architecture discussed later) weighing 1 kg apiece. They could save 90% of the cost of the originally planned 1,000 kg payload. This could save considerable money for taxpayers, and accomplish vastly more in the way of Mars exploration. According to Brooks (F&M, p. 56), an engineer at the Jet Propulsion Lab said that the government “Wanted a big instrument, not a little bitty one,” and it would be too much of a let down for his co-workers who had waited fifteen years to accept anything less than this.

Well, I guess that's government as usual for you. But one might ask, "What was so `fast, cheap, and out of control' about the Roomba when it was introduced in 2002?"

As for “cheap,” it originally retailed below $200. This is considered a major price point for the consumer in the novelty-gadget market.

As for “out of control,” in order to bring the costs down, iRobot could not afford to install an expensive navigation and sensor system. A compulsive need to do this was a hang up of many competitors. Instead, the Roomba runs on a sweeping motion algorithm originally devised to clear minefields. The robot does not need to systematically run back in forth in non-overlapping rows, Nor does it need to have an internal map of an area it covers. It does, however, have sensors necessary to avoid falling down stairs and to clean around furniture without getting stuck. A“virtual wall” accessory prevents the Roomba from straying out of a room or hallway.

As for “fast,” the Roomba is in fact slower than a human performing an equivalent vacuum cleaning job. However, it is “faster” when we redefine the job concept from a robotic perspective. A human who might spend twenty minutes vacuum cleaning a room by himself now only has to spend a couple of minutes to turn on the Roomba. The Roomba itself might take over an hour to do the job, but that is not the human’s time problem anymore. A Roomba sales slogan is: “Turn it on and walk away.”

The Roomba debuted in three stores in 2002. Within a year the number of stores increased fifteen fold. At 21-23 Oct 2004 RoboNexus conference, Helen Greiner announced that iRobot had just past the 1,000,000 unit point and it is now sold in over 7,000 retail locations. Colin Angle showed a time-phased sales volume graph that he overlaid with a graph of Apple Computer's iPod sales to show an interesting similarity.

“They’re the first ones to make a commercial success of what would be considered domestic robotics,” said Craig Jennings, president of the Robotic Industries Association. “Nobody else has had a product that has had the success of Roomba. I think they hit a home run.”

iRobot has now come out with its "next generation" line-up. Its latest models hover longer over dirty areas, offer superior cleaning power, return by themselves to recharge docking stations, require substantially less charging time, and show many other improvements.



The PackBot is now a “battle-tested” success. The Future Combat Systems (FCS) Small Unmanned Ground Vehicle (SUGV) depicted to the right is currently in development. [credit: iRobot Corp.]

The May 27, 2002 Forbes article “Machine Dreams” describes how in 1993 iRobot got its first major steady income with a $50,000 grant from the Defense Advanced Research Projects Agency (DARPA) to build a robot crab named Ariel to handle mines for the Navy. Then in 1995, following the Oklahoma City bombing, iRobot got a $3 million contract to build the PackBot reconnaissance robo-tank. iRobot took the PackBot from the concept stage to combat in 3.5 years, and has adapted its joystick/wired-guided platform to perform a variety of missions, to include carrying an Explosive Ordnance Disposal manipulator arm or a surveillance camera.

A number of fire fighters balked at using the PackBot at the 9/11 World Trade Center site for fear that if they lost a robot in the rubble, they would have to go in after it. In contrast, many Army soldiers who balked at the idea of using the PackBot while in urban training in the U.S., suddenly changed their tune when faced with the prospect of getting their heads blown off in caves in Afghanistan. The Army has been hooked on the PackBot ever since, and iRobot can’t build them fast enough. They range in price from $50,000 to $108,000 apiece. For the encore, the Army has awarded iRobot a $32 million contract to develop a smaller version of the PackBot called the SUGV that weighs under 30 lbs and can be carried by a soldier in his backpack.



Machine Dreams" described other iRobot projects. “Meantime, Angle and Greiner finagled a contract with Baker Hughes, the oil service company. The result was a titanium-encased, 80-centimeter wireless robot [MicroRig -depicted to the left inside an oil rig with two chain propulsion units running parallel on each side of the casing -author]] that could go eight kilometers down an oil-well bore to make repairs; it could carry loads and withstand a temperature of 150 degrees Celsius and a pressure of 700 kilograms per square centimeter. (Baker Hughes backed out; Halliburton has since picked up an interest in the device.)... " Twenty out of the thirty foot length of this "robot worm" are devoted to batteries.



"Machine Dreams" continued, "Even more promising is an adult robot called CoWorker, a meter-tall device with a video camera that can transmit live images from remote locations. Moving at a kilometer and a half an hour, this mobile machine is remarkably nimble, turning sharp corners, negotiating tight spaces. Disney and Halliburton are testing the bot, which cost $2.5 million to develop. Someday, perhaps, the CoWorker might serve as a sitter for the elderly or help a doctor by making his house calls for him. “


It is harder than it looks: Colin Angle pointed out during his talk at the 9-10 March 2004 robotics conference that robots tend to be complex. A normal vacuum cleaner has 50-100 parts. Roomba has 1,000 parts and 1,000 lines of code. The PackBot has 10,235 parts in 409 subassemblies. This makes it comparable to an automobile in complexity.

In some ways he felt like iRobot is a cross between Apple Computer and Ford. However, in regard to the Ford comparison, cars are certainly not as "ruggedized." If you drop a car from more than its body length, it will likely break. In contrast, the PackBot is designed to survive 400+ G’s, or the impact of being thrown out of a second story window.

Mr. Angle added that a corollary of robot complexity is that it is still very hard to find a commercial application that justifies the costs of robot production. In order to find each profitable application, he has had to explore innumerable project ideas, and has experienced many disappointing ventures.

My Real Baby seemed exciting at first, but it ended being a much more high risk venture than originally expected. It entailed considerable risk because the doll retails around $99. High tech dolls in general cost $80 to $120. This is certainly on the high end of what parents will pay to get dolls for their daughters.

According to Mr. Angle, all of this puts stress in three directions. One is to focus more time on finding special market niches and well thought out business plans that can justify current robot development and production costs. Secondly, once volume increases to a certain level, it also suggests opportunities further upstream in the “structure of production” for original equipment manufacturers to drive down costs by focusing on increasing efficiencies in making subassemblies, accessories, and maintenance kits. Last, but not least, we always have a waiting game in robotics for prices to continue to come down further as capabilities rise. We also need to see new technologies reduce complexity as well as costs further in each of the robotic component areas.

At the 21-23 2004 Oct RoboNexus conference, Mr. Angle told me that the big issue with iRobot right now is focus. They have Roomba and PackBot which are both very profitable, and they want to stay focused on them. Yes, they still hold the patents for MicroRig. They still believe in the remote presence concept behind CoWorker,. However, they are concerned about spreading themselves too thin, particularly in areas where off the shelf component costs still have not come down to the point that the business opportunity becomes irresistible. A big question with CoWorker is what features need to be added (e.g. more versatile propulsion system, two-way video-conferencing or hand manipulation), and then to what price point must CoWorker be brought down to before it becomes a "killer app." Apparently iRobot has backed away from the idea that the robot could be marketed as a very high end $3,500 "Telefriend" novelty similar to the Aibo. The latter may be completely different novelty case involving established demand substitution, to the extent that in Japan many apartment complexes forbid or charge high fees for living pets.

All the old issues that beset vertically integrated manufacturing firms also beset iRobot's Roomba. Helen Greiner pointed out that iRobot's production costs are over 20% of total budget. To keep costs down, the Roombas are manufactured in China and the PackBots are made in Hong Kong. Manufacturing costs are so critical that they like to separate their manufacturing and design teams as soon as possible in project development. She noted that we are still a long ways from the day when robots will manufacture other robots.

According to Ms. Greiner, while "quality control" is a phrase they did not use fourteen years ago as an MIT spin-off, today it is a major concern. A consumer robot not only has to have low "intellectual load" (i.e. be easy to use), but also be nearly 100% reliable (reduce consumer frustration). After the debut of the first generation Roomba, they discovered that many users wanted to use it once a day rather than once a week, so they had to make design changes to meet their goal of a three to four year longevity.

As an example of a quality control problem, in regard to a first generation Roomba model, customers complained about a battery problem. They said that when the robot sat on a charger for a week, it discharged completely. iRobot discovered it had to change a resistor value to alter voltage stress. This not only meant fixing 100,000 robots sitting on store shelves, but also spreading the word with Internet user groups to alert current owners. The problem cost iRobot $750,000 to fix. The company was nevertheless pleased with the good-natured response of its customers for honestly and quickly addressing the problem. If a Roomba is defective, iRobot replaces it to prove that they stand behind their products and build their brand.

Quality concerns also involve the end effector. Quite often in robotics the quality of the end effector that does the actual work is more important than the underlying robot itself. According to Colin Angle, the second generation Roomba has twice the suction power compared to the first generation. The independent research firm Intertek claims the Rooma now has more cleaning power than most conventional vacuum cleaners. To accomplish this feat, the Roomba uses two counter rotating brushes that scoop up as much debris as possible by mechanical action. It employs suction through a thin long slit between the brushes that has a much higher air velocity than convention cleaners. As Mr. Angle put it, the Roomba reflects a fresh look at the old vacuum cleaning physics.

iRobot also has to deal with distribution issues. They found that when retail stores independently display and explain the Roomba to prospective buyers, it sells at double the rate compared to when it is left bundled on a retail shelf with other consumer products. Overcoming consumer skepticism about this new product is a major issue. Television ads that show the Roomba in action have been particularly effective to demonstrate that it can move as quickly as demonstrated. One campaign sold 10,000 Roombas in 2.5 hours, generating $2 million in revenues.

Like many tech firms, iRobot has to deal with proprietary vs. collaborative business issues. iRobot has created a serial port on its Roomba that allows programmers to access it software. It will make API code available on the Internet to encourage entrepreneurs to write innovative software. iRobot also found with its PackBot in Afghanistan and Iraq that it needed to increase the ability of military users to access and customize its software in the field.

Last, but not least, iRobot has to deal with direct competition. Evolution Robotics has developed its e-Vac robotic vacuum cleaner priced at $199.99 in the Sharper Image catalog. This robot senses the dimensions of a room to engage in "deterministic" navigation that allows systematic sweeping. An Evolution employee told me that their robot is selling at the same rate out of the catalog as the Roomba. This is not surprising, since the e-Vac appears earlier in the catalog, and the page exposure and list of features for each probably look similar to people who are unfamiliar with robotic vacuum cleaners. As another example of potential competition, the French firm WANY robotics displayed a robo vacuum cleaner with deterministic navigation at Robo Nexus that it believes can also price below $200.

iRobot has responded to competitors in its second generation product line by offering models that range from the Roomba Red at $149.99 to its Roomba Discover that retails at $249.99. Its second generation robots also offer twice the cleaning power and three times the dust bin capacity compared to the first generation. A former iRobot employee told me that he ran tests and found that iRobot's semi-random algorithm showed better cleaning performance compared to Evolution's deterministic navigation under certain conditions, such as in rooms where people and objects move around or have other peculiarities. Conversely, there are other conditions where the deterministic approach is quicker.

My guess is that because sensor and navigation software costs are steadily declining in costs, iRobot will eventually offer a deterministic navigation option in a next generation model to fend off competitors and defend its first mover advantage and dominant market share. It will offer "fast, cheap, and in-control." However, this may not be a big issue, since most people may not care that a semi-random robo vacuum cleaner tends to overclean a bit and take a little longer in return for more flexibility.

Obviously there are many pieces to the business jigsaw puzzle one must consider to have a successful robot company. More on all of this in Part Four, my section about making money in mobile robotics.

Link to
......Part III of series

Jump back to...... Part One.....

Jump forward to...... Part Four...... Part Five
.......Part Six

Disclaimer: This report is for research/informational purposes only, and should not be construed as a recommendation of any security. Information contained herein has been compiled from sources believed to be reliable. There is however, no guarantee of its accuracy or completeness.

Bill Fox is VP/Investment Strategist, America First Trust. Bill welcomes phone calls and email responses to this article. His most current contact information is at his web site: www.amfir.com.

Short URL for this web page: http://tinyurl.com/2425tj8


Flag carried by the 3rd Maryland Regiment at the Battle of Cowpens, S. Carolina, 1781

© William Fox. Sometimes William Fox offers viewpoints that are not necessarily his own to provide additional perspectives.