Arkansas shoppers who like ordering their groceries online and picking them up at their local Walmart will soon get a bit of extra help from a self-driving robot—though few will realize it. In a deal announced this week, Walmart will use robo-vans from startup Gatik to help move goods from its Supercenter in Rogers, Arkansas, to a Neighborhood Market in nearby Bentonville, where customers can get their bags with laundry detergent, paper towels, potato chips, and whatever else.

In the pilot program, Palo Alto, California-based Gatik will run three of its modified Ford Transit Connect vans in this “middle mile” logistics scheme, making up to 10 runs a day, seven days a week, during daylight. With human safety operators behind the wheel as backups, the vans will follow one of two routes between the stores: a roundabout five-mile (one-way) trip that avoids some tricky driving scenarios, or a more direct, uncompromising two-mile shot. “The main aim is to do that repeatedly, safely, without a safety driver onboard,” says Gatik CEO Gautam Narang.

The difficulty of that task—operating reliably enough to eliminate the human backup—explains Gatik’s limited ambitions. The startup has no plans to move on to robo-taxis or deal with individual passengers or even recipients of whatever it’s moving. “We don’t have to wait for changing consumer behavior,” Narang says, meaning waiting for people to become comfortable with the idea of riding inside a robot, or even walking to the curb to retrieve their order from a robot that can’t walk up to their door. Gatik wants to work with businesses like Walmart, finding the spots where cutting out the cost of human labor makes a robot valuable. Narang says Gatik can cut the cost of the trips it’s making for Walmart by as much as 50%. Gatik declined to say how long the pilot will last, but a spokesperson said that if all goes well, the companies may extend their relationship to other routes.

That only works, though, if its vehicles can handle the routes on their own. Using the standard suite of lidar, radar, and camera sensors, Gatik’s van will face relatively complex scenarios, including lane changes and an unprotected left turn—an enduring bugaboo of robo-car developers—on the shorter, more direct route between the stores. During rush hour or other busy times, when moving fast isn’t a priority, it can use the 5-mile route, which takes a bit longer but avoids the tricky left. (Narang points out that UPS human drivers don’t turn left much either, but that’s more a question of fuel economy than ability.)

Funded by a $4.5 million seed round led by Innovation Endeavors, Narang’s team of 15 to 20 people is spread among tech centers in Palo Alto and Toronto, along with an operations team in Arkansas. As with other startups joining the robo-fray in recent years, Gatik has a very specific target in mind. It wants to sit between the long-haul trucks of Kodiak, Pronto, Starsky, and Ike, and the last-mile delivery work of Nuro, Boxbot, Refraction, and sidewalk robots shops like Marble and Starship.

These limited targets are easy to understand when looking at the trouble that the biggest players in this sector have had delivering on the far more ambitious goal of using autonomous vehicles to move people. Earlier this week, General Motors’ Cruise announced that it will miss its deadline of deploying its robo-taxi program in San Francisco this year. Waymo made its 2018 deadline for launching its service but only by keeping its human backups behind the wheel.

So it makes sense for Gatik, about one-hundredth the size of Cruise, to find a smaller role to play. And however well its technology works in Arkansas, it serves as one more sign of how autonomous tech will make its way into the world—not in one fell swoop, but bit by bit, in places you might not think to look.

This story originally appeared on wired.com.

https://arstechnica.com/?p=1542129