Simbe Robotics today announced adding RFID scanning and machine learning capabilities to its retail inventory-scanning robot, Tally. With the new technologies, Tally can now capture data from more than 700 RFID tags per second with higher than 99% accuracy, the company said.

Tally, which debuted in late 2015, already includes computer vision capabilities for scanning store shelves to determine out-of-stock or misplaced items for retailers such as grocery stores and pharmacies. With RFID scanning, the robot can now determine inventory levels for products that utilize RFID or that don’t sit on a standard grocery store shelf.

“Through the announcement of that product, we received an immense amount of interest from other types of retailers, particularly those that focused on high-value products, whether it’s clothing or soft lines, sporting goods, and consumer electronics,” said Brad Bogolea, CEO and co-founder of Simbe Robotics. “So it became clear to us that adding RFID capabilities to Tally would open us up to a larger swath of the market.”

The Tally robots have been deployed at 11 retailers worldwide, taking more than 32 million shelf photos, analyzing more than 150 million products and shelf tags, and navigating more than 4,500 kilometers inside stores alongside customers and employees, the company said. Out of those 11, three of the retailers are also using RFID scanning to capture data, with more than 150,000 products scanned over the past nine months.

“One of the key components here is that we’re the first robot to combine both computer vision and RFID into a single solution that’s really designed to operate in store environments,” Bogolea said.

The Tally system is designed to help retailers with improving the frequency of inventory audits, which human workers usually only have time to do once a week or every other week. In Simbe’s trials, Bogolea said retailers “have moved to a model where Tally is often either perpetually scanning, or at least performing one site-wide inventory audit per day.”

In addition, the system helps retailers determine the precise location of items, especially when customers do things like move them around in store. “Often, shoppers have a tendency to pick things up and put them down somewhere else,” Bogolea said. “So understanding not only what the inventory count is in the store, but precisely where those tags are, is key [for retailers].”

Machine learning capabilities on the robot help it learn more about the paths it takes for navigating through a store, around customers, and avoiding obstacles, Bogolea said. Algorithms are also used to help Tally improve its accuracy when scanning for the RFID tags, or using the computer vision capabilities to take photos of store shelves.

Simbe said data from the robots is sent securely to the cloud for processing and analysis, and then sent back to key stakeholders through an API and front-end application, with the goal of improving store performance.

Simbe said its goal is to help retailers solve the $1.1 trillion in annual losses that retailers face due to products being out of stock, merchandise overstock and product location errors.

The San Francisco-based company is privately held backed by venture companies, and Bogolea has said they have remained quiet about their funding, which has included several undisclosed seed rounds and angel investments. In April, the company announced a global partnership with SoftBank Robotics America to expand deployments of Tally in Japan, North America, and Europe.

The Tally robots are similar in style to those developed by Bossa Nova Robotics, which recently announced raising $29 million to scale up its mobile inventory robots.