Walmart to test drive shelf-scanning robots
Robots aren’t just for sale in the toy aisle, they’re gliding down them.
Strip District-based Bossa Nova Robotics Inc. is sending its shelf-scanning robots out to 50 Walmart stores in a real-world use of technology to help one of the planet’s largest retailers keep its aisles stocked and ready for customers.
The “Bossa Nova Robot” uses sensors similar to those on selfdriving cars to navigate — taking photos of shelves, as well as recording data about products’ prices, locations and if they’re out of stock.
The tiny white robot — about 2 feet tall without its camera tower — isn’t expected to replace
replace workers. It simply scans, passes information to the cloud, communicates that data to Walmart’s backend system and relays that knowledge to store associates.
“It’s a tool that lets them focus on how to improve the shopping experience,” said Sarjoun Skaff, chief technology officer at Bossa Nova. “It helps them prioritize what to do.”
In October, the retail giant told Reuters it was deploying shelf-scanning robots in California, Pennsylvania and Arkansas.
While Walmart would not yet give specifics on stores in Pennsylvania, the company did confirm it has a relationship with Bossa Nova Robotics, which employs 89 people between two offices in Pittsburgh and in San Francisco.
“We have worked closely with Bossa Nova to help ensure this technology, which is designed to capture and share in-store data with our associates in near real time, works in our unique store environment,” said John Crecelius, vice president of central operations at Walmart.
“This is meant to be a tool that helps our associates quickly identify where they can make the biggest difference for our customers.”
Capturing lost customers
With e-commerce sales growing exponentially each year, brick-and-mortar chains need to operate near perfectly to retain customers.
In August, the Census Bureau estimated digital sales in the second quarter for the 2017 fiscal year would increase 4.5 percent to $111.5 billion. Total retail sales are estimated to increase just 0.5 percent from the first quarter, reaching $1,256.2 billion.
A lost sale in a physical retail store always has been an inefficiency, but now the stakes are higher.
Bossa Nova has been testing its robots since 2013, registering more than 710 miles and 2,350 hours of autonomous inventory scanning, capturing over 80 million product images.
“The information [the robot] is capturing is whether there’s an out-of-stock, because that is the biggest frustration of shoppers,” Mr. Skaff said. “If they don’t have it, you’re going home without it — or maybe you go to a different store.”
Mr. Skaff said that in the pilots, no customers have been spooked by the little R2-D2-esque machine.
“Either they’re like, ‘Oh my God, this is so cool. Science fiction has arrived,’ or they ignore it as if they’ve been seeing robots all their lives,” Mr. Skaff said.
That reaction, or lack thereof, has been precisely engineered. Mr. Skaff said the robot’s design must strike a delicate balance.
“If it’s too friendly, [the customers] may want to interact with it and not let it do its job, and if it is too much of a tool, people may simply not like it.”
‘A slow self-driving car’
The Bossa Nova Robot employs a number of sensors, including light detection and ranging (lidar) technology, a critical element in self-driving cars. On a self-driving Uber, that’s the revolving piece on top of the hood, spinning to capture the world in 3D.
Mr. Skaff said his team competes with the self-driving car industry for talent, which makes retail robotics “a very hard market.”
Gliding slowly down a store aisle — at about 0.4 meters per second, or a rather lax walking pace — the robot scans its environment with depth sensors to avoid collisions with shelves, or more importantly, people.
“It’s a slow self-driving car, if you want,” said Mr. Skaff, with a laugh.
Inside an unmarked testing space across the street from Bossa Nova’s headquarters at the Cigar Factory on Smallman Street, Mr. Skaff explained how the robot works.
As it scans its environment, the robot is collecting data with a tower of cameras, set up along the length of a pole that reaches a little over 6 feet high. The cameras only point at the shelf to capture product images. Bright white lights illuminate the shelf for better image quality.
As soon as a customer gets close, the robot moves out of the way and shuts off its high beams.
Since the cameras can only really see the front of a shelf, the inventory data is binary. There are only two options: stocked or not stocked. For now, it can’t record the exact figures for products on the sales floor.
For a similar reason, it’s not outfitted for use in the apparel section. It’s difficult to see the tags on clothing, since they’re sometimes hung on a rack, folded on tables or placed high up on a wall.
Robotic competition
Despite the sheer size of Bossa Nova’s Walmart pilot — not to mention the company’s at least $23.1 million in backing as selfreported to funding platform Crunchbase — it’s not a lone wolf in retail robotics.
Minneapolis-based Target tested an inventory checking robot at a San Francisco store in 2016.
San Francisco-based Simbe Robotics designed the robot, called Tally, as an autonomous shelf auditing and analytics tool — identifying and monitoring inventory levels for out-of-stocks, mispriced items and low-priced items.
Lowe’s, the Mooresville, N.C.-based home improvement retail store, has taken a slightly different approach. Fellow Robots, based in Mountain View, Calif., helped the company create LoweBot, which the retailer revealed in San Francisco stores in 2016.
LoweBot has a computer for a face, which helps customers to find products in a store or to ask basic customer service questions by touching the screen or audibly asking for assistance, all while collecting inventory data.
Mr. Skaff remains unconcerned with the competition. Inventory issues have been a vexing operational issue for retailers across the board, he said, which means Bossa Nova will naturally have competitors.
“The beautiful thing about a robot is that it is not a big infrastructural installation,” he said.
That means that a successful pilot could scale and allow the company to eventually license its robot to other big box stores.
“It’s been unanimously super exciting. It’s very rewarding,” Mr. Skaff said. “We go and deploy the robots, and we start to hear from the associates themselves, how excited they are about using it as a tool.”
“Either they’re like, ‘Oh my God, this is so cool. Science fiction has arrived,’ or they ignore it as if they’ve been seeing robots all their lives.” — Sarjoun Skaff, Bossa Nova