Priya Narasimhan has a vision: a robot helping shoppers find everything on their shopping list in stores such as Giant Eagle or Costco. Walking through the Carnegie Mellon campus bookstore on a recent Wednesday, she was accompanied by just that -- a 31/2-foot robot resembling R2D2.
It's the latest project to come out of the university's robotics institute. Ms. Narasimhan, a CMU professor and businesswoman, invented AndyVision, a robot, digital screen and information software that together enable communications among customers, the retailer and the robot.
Ms. Narasimhan said the idea for AndyVision, funded by CMU's Intel Science and Technology Center, came out of her own frustrations with shopping in grocery stores.
"You go into stores and you get frustrated because you can't find what you need or you leave feeling shortchanged because you bought all the wrong things," she said.
The current prototype of the robot, known affectionately as ScotBot, can navigate autonomously on wheels around the store, taking photos with its high-resolution camera. It recognizes items based on their shape, color and location, and its visual-recognition technology notices when something is missing, misplaced or low in stock. Eventually, its makers plan to link the robot with store owners' mobile devices or laptops, providing them with real-time inventory updates.
Customers also can browse a large touch screen at the front of the CMU store. They can see a 3-D image of merchandise that is on sale and scan a QR code from the screen onto their mobile phones to get the discount. Ms. Narasimhan also wants to use a three-dimensional store image to help shoppers map the location of every ingredient they need to bake a cake, or see where the guacamole is when they buy tortilla chips.
Like many roboticists, Ms. Narasimhan must balance a dream of robotics with smart business. Though she said she wants retailers' needs to determine which technologies she incorporates into AndyVision, it could turn out to be yet another endeavor whose software sells better than the robot itself. Before it can reach the marketplace, major costs, time and research barriers remain.
CMU's Intel has spent $200,000 on the project to date, helping Ms. Narasimhan's team buy materials and paying the salaries of four graduate students and post-doctoral fellows. The science center has devoted at least $2.5 million a year to various projects since 2011, part of a five-year commitment. This fall, Ms. Narasimhan will pilot ScotBot by letting it work alone in the CMU campus bookstore. If that's successful, she hopes to run pilot programs at retailers such as Costco and grocery stores over the next year, and to sell the system to retailers after that.
Unlike in the defense robotics industry, there are no research and development contracts available for retail robotics. Researchers must demonstrate that a robot can perform in other environments before retailers will even consider it, said Paolo Pirjanian, CEO of Evolution Robotics in California.
"The retail industry are not early adopters. They're laggers when it comes to this technology," he said. But once one retailer signs on for new tech, "then everyone wants to jump on."
Retailers do strive to keep up with the latest technological improvements to best serve their customers, Kathy Grannis, senior director of media relations at the National Retail Federation, said in an email. But she could not speculate on how many would embrace something like AndyVision because it's not currently being used in stores.
Consultants from Boston Retail Partners said if a technology can help save on labor costs, as Ms. Narasimhan hopes her robot will, it can make a difference for retailers. Labor is the second-highest controllable cost for grocers, after refrigeration, they said. Much of that labor is devoted to counting goods several times a day. But Perry Kramer, vice president of the consulting firm, said grocers have some of the lowest profit margins in retail, making them cautious about adopting new technologies.
Mr. Kramer and Brian Brunk, a partner at the consulting group, both said AndyVision's software could help improve customer service by linking information about the location of goods with customers' cell phones. But buying the AndyVision software without the robot might be more attractive to some retailers, who prefer to use known, cheaper technologies, they said.
Ms. Narasimhan, whose two previous robotics ventures succeeded as companies (she is founder and CEO of YinzCam, which enhances viewing of sports events on mobile devices), said she doesn't know yet if retailers will want to buy AndyVision with ScotBot or just rely on cameras placed on shelves or carts. That kind of technology is already ubiquitous: The California company Datalogic ADC, for example, mounts cameras in checkout lines and on shelves in some groceries to track items missed at checkout.
She said the robot's roaming capabilities provide retailers with the added advantage of constant, real-time information about inventory.
"We are enamored and attached to the concept of the robot," said Ms. Narasimhan. "But from a very practical standpoint, we realize it may not be feasible everywhere."
Martin Hitch, CEO of Bossa Nova Robotics, said if robots succeed in retail, they'll find a path into homes and into the mainstream as their cost slides.
"I don't think there's a lack of personal robots because there's a lack of ideas," he said. "I think there's a lack of personal robots because the cost is prohibitive."
Catherine Mott, CEO of BlueTree Allied Angels in Pittsburgh, an angel-investing company, said the robotics industry's long-term returns make investment less attractive. Angel investors expect 10-times growth on investment in five years. Venture capitalists can't wait longer than 10 years for their money to grow. That means they're more likely to invest in software than in research to advance robots with longer-term projections for success.
"What's unfortunate is that some of the large, game-changing technologies require so much capital for proof of concept that it's harder to get them to the marketplace," she said. "Many robotics companies are a technology looking to find a problem instead of starting with a problem that they can solve."
But Ms. Narasimhan knows what the problem is -- she's lived it. The question is whether her team has given the problem more technology than it needs to be solved.
She said it's possible that if Intel doesn't continue funding research to perfect ScotBot, she will sell the software directly to retailers, allowing them to make the choice about whether to buy a robot or just cameras.
Seema Patel, CEO of the Pittsburgh company InterBots, which makes robots to assist children with autism, thinks robots will remain confined to niche markets -- at least for the foreseeable future.
Jim Ostrowski, vice president of product development for Datalogic, guessed it would take 10 to 20 years to see robots in stores and homes.
For now, it's products like iRobot's $300 automated vacuum cleaner that are selling big -- 6 million units to date.
Ms. Patel said that might be as far as robotics goes: "I don't know that we'll ever have Rosie Jetson, but we might have a really smart dishwasher or self-cleaning toilets."
Still, many roboticists are hopeful that ScotBot and friends will find their way into homes, malls, schools and hospitals one day.
And Mr. Ostrowski said he still thinks the dream of a C3PO-filled world is worth aspiring to -- because one day, robots may do more than count inventory. They might stock the shelves or carry crates, eliminating the "dull and dirty" tasks from daily life.
Or they may already be here.
Sanjena Sathian: firstname.lastname@example.org or 412-263-1408. First Published July 29, 2012 4:00 AM