Mowing on Autopilot

Udi Peless dislikes lawn mowing. Strange then that Peless — who has made his mark as an F-16 pilot and a highly successful entrepreneur — should now be living a life that revolves around lawn mowers. And it's even stranger that Peless is doing so in Israel, a land not known for expanses of abundant, green, residential lawns.

But Peless, who holds a bachelor's degree in electrical engineering, also happens to know a thing or two about navigation, which is why his vision of landscaping involves lawn mowers on autopilot. Because of his distaste for mowing, Peless spent the past decade working with his partner, Shai Abramson, on the design of a stubby lawn mower that combines electric motors, sensors, software and a microprocessor to automate the process of grass cutting. Their start-up company's high-end RL1000 RoboMower, which recently began to sell through Sears.com in the U.S., is so autonomous it can mow for an entire summer without a shred of human intervention, docking and undocking itself from its battery charger, and independently finding its way around the yard each week.

Peless and Abramson, along with their U.S. distributor, believe such mowing systems will emerge as a must-have home automation technology in the next decade, much as garage door openers did in the 1960s.

"Some day, when you buy a house, you'll say, 'I don't want to buy a house without a mowing system,'" says Ames Tiedeman, national sales manager for Systems Trading Corp., the U.S. distributor for the RoboMower.

Indeed, others have shared that vision. To date, the company has sold approximately 50,000 lawn mowers, about one-third of those in the U.S. Moreover, major manufacturers — including Toro, John Deere, Stiga and even Hoover — have had discussions and/or business agreements with Friendly Robotics, the start-up company launched by Peless and Abramson.

"We've been received with genuine interest and a lot of curiosity," Peless says. "Clearly, people see value in the concept."

Sensing challenge

That, however, was hardly the case when Peless first cooked up the idea in 1995. But he would not be deterred by skeptics, mainly because he believed there was a market for robotic lawn mowers.

"As a consumer and a homeowner, I hated mowing my lawn," he recalls. "I wanted to develop something that would do it for me, and for all the other homeowners who felt the same way."

Peless, who already enjoyed huge success with an earlier medical start-up, known as Biosense, saw the robotic lawn mower idea as another start-up with great potential. Based on his assumption that there was a need in the marketplace, and bringing his own experience in the design of navigation and control systems to bear on the project, he forged ahead with the idea that robotic lawn mowers were both technically feasible and sellable. He contacted Abramson, who he knew possessed expertise in electro-optics and software, and together the two began work on a garage prototype. It wasn't long, however, before they realized that the project was more difficult than either had anticipated.

"Unfortunately, we were overconfident," Peless recalls. "It took several years and several million dollars more than we expected."

Still, the basic concept for the robotic lawn mower took shape early, and with minor exceptions, has remained mostly intact. The pair's original idea called for the mower to employ a variety of sensing systems, including an ultrasonic proximity sensor, an electromagnetic field sensor, and so-called "bumper sensor," as well as an inexpensive odometer and magnetic compass.

During operation, the mower's most important sensing element was its electromagnetic field sensor, designed to work in conjunction with a wire laid around the perimeter of the property to be mowed. In essence, the system worked like a transformer, using a primary coil outside the mower to generate a low-voltage signal in the perimeter wire, and employing a secondary coil on-board the mower to sense its distance from the wire. By working in conjunction with the odometer and compass, the system could also gauge how far it had gone, as well as its heading. At the same time, unexpected obstacles, such as rocks or passersby, could be "seen" by the prox sensor and "felt" by the bumper sensor, which used an on-board pneumatic tube and pressure sensor to detect the slightest impact in its path.

To ferret through the input from all the sensors, Friendly Robotics fed all the sensor signals to a printed circuit board that employed a 16-bit, 20-MHz Hitachi HS8 microcontroller to weigh the inputs and send command signals to the drive system and the cutting blade motors. Peless and Abramson initially used electric motors from Johnson Electric (Hong Kong): one 150W motor at each of two drive wheels; and a bulky 750W motor to power the cutting blade.

"By controlling the drive motors, we were able to navigate the unit and control its speed," says Abramson, a longtime engineer from the Israeli army who holds degrees in physics, mathematics and electrical engineering. "Using differential speed between the motors, we could turn and navigate at the same time."

Walking the line

Peless and Abramson, however, agree it was the software, not the hardware, that provided engineers with their greatest challenge.

"The real breakthrough was the use of the compass and the odometer together with the signals from the perimeter wire to determine where we were," Abramson explains. "It's a low-cost navigation system working together with a coverage algorithm that forms the heart of our system."

Indeed, the combination of technologies enables the mower to travel in straight, precise lines that form slightly overlapping coverage stripes on the lawn. By working in this way, the system leaves no area uncut, even while traversing a bumpy, hilly course.

"These are lawns, not tennis courts," Abramson says. "Keeping straight lines on bumpy, hilly ground is a tremendous challenge."

Indeed, it took a software group of six engineers approximately 20 man-years of time to perfect the coverage algorithm, Abramson says.

"In a simple garden, you could navigate your way across from side to side, and it wouldn't be difficult," he says. "But if you have a more complicated garden with obstacles and narrow passages, the system has to trigger specific behaviors to help it deal with the particular shape. It took us a long time to write the software because it required so many hours of testing and collecting data for every possible shape."

"It was pretty easy to cut 80 percent of any lawn," adds Peless. "But to get 100 percent coverage for every possible shape took years of perfecting the algorithm."

Going to market

Still, Friendly Robotics managed to bring its first product to market by 1998. Although its founders now acknowledge that first version was bulky and lacked the robustness of later models, they say its performance was impressive for the time.

"It became obvious pretty quickly that this was a good first shot, but not a product we could build a lasting business on," Peless says.

The company built about 4,000 of the units between 1998 and 2001, before launching second and third versions that employed a variety of improvements. Those new versions used a vacuum forming process to create less costly plastic molds for the mower and swapped out the original 750W cutting motor for three smaller 150W units that powered three smaller cutting blades. The mower's design team also dropped the proximity sensor from later units, and added an optical "drop-off sensor" that enabled the RoboMower to understand if it was tilting, or if one wheel had left the ground. In such cases, the microcontroller would shut down the cutting blade motor as a way of preventing injury. Software engineers also changed the coverage algorithm to create a zig-zag pattern across the grass.

"With the first pass, we now cover 70 to 80 percent of the lawn, and then we switch and cover the garden from a different angle, creating parallel overlapping stripes," Abramson says.

While developing the RoboMower, Peless and Abramson have also negotiated with major manufacturers in an effort to license their technology to a company with greater manufacturing wherewithal. Friendly Robotics struck up a formal relationship with Sweden-based lawn mower manufacturer Stiga and with U.S.-based Toro. Toro even produced and sold several hundred robotic mowers under the trade name iMo, but eventually dropped the idea for lack of customer interest. Peless also negotiated with Hoover Co. in an attempt to license the technology for use in robotic vacuum cleaners, but negotiations between the two firms ended after Hoover was acquired by Maytag Corp.

"Unfortunately, we were disappointed in all three cases," Peless says. "So we eventually ended up making and selling the product under our own brand."

Friendly Robotics has since managed to sell more than 50,000 mowers, largely in Europe and in the U.S., and it has committed to making the product by itself, at least for now. When market interest in robotic mowing rises, the company's principals say they'll reconsider licensing.

Until that time, distributors of the RL1000 and the smaller RL850 say they're hoping for big sales volumes.

"There's 6 million lawn mowers sold every year in the U.S.," notes Tiedeman of U.S.-based Systems Trading Corp. "And our goal is to get 10 percent of that market."

Peless and Abramson contend that volume will help drive down the price of the RL1000, which now retails at $1,800, and the RL850, which retails at $1,200.

Whether or not the start-up is able to boost its sales figures soon, its founders believe they have opened a new door in the fledgling home robotics market.

"This is a whole new category," Abramson says. "It's not like making televisions or refrigerators. There was a psychological gap that had to be crossed here and now that consumers have begun to cross it, they won't go back."