Amazon-backed RIVR is testing delivery robots for extreme weather

Signals Inbox·July 26, 2026·Robotics

Amazon is putting RIVR’s delivery robots through snow, rough ground, stairs, and other conditions that usually expose the gap between a good robot demo and a useful delivery machine. The bigger tension is that Amazon already tried this once with Scout. It ended those field tests in 2022, then acquired RIVR in March 2026 for another attempt built around wheels, legs, and the difficult final steps to the customer’s door.

Q1What is RIVR actually testing?

RIVR’s official product page says its robots are built to cover the last mile and the final 100 yards to the doorstep. The latest tests push them through bad weather and difficult ground to see whether they can keep moving when sidewalks become wet, snowy, uneven, or blocked. This is reliability testing for the messy world outside a lab.

Q2Why does Amazon care about rough terrain?

Because getting close to a home is not the same as reaching its door. A basic sidewalk robot can stop at curbs, steps, gravel, gates, or steep paths. Humans handle those obstacles without thinking, but robots regularly fail there. RIVR uses wheels for speed and legs for obstacles, trying to automate the part of delivery that has remained stubbornly human.

Q3Has Amazon tried delivery robots before?

Yes. Amazon began testing its six-wheeled Scout robot in 2019, but ended the field tests in 2022 after saying parts of the program did not meet customer needs. Scout mainly travelled on sidewalks. RIVR is a second attempt with a more flexible body that can climb stairs and move across terrain that would stop a normal wheeled machine.

Q4Is RIVR already making real deliveries?

Yes, but at trial scale. RIVR has tested parcel delivery in Austin and worked with Just Eat on food deliveries in Milton Keynes. A previous Swiss program completed close to 1,000 deliveries, while the UK launch involved ten restaurant brands. Those trials prove the robot can work outside. They do not yet prove it can operate cheaply and reliably across Amazon’s huge network.

Q5Why test extreme weather now?

Amazon acquired RIVR in March 2026, so the question has shifted from whether the robot looks impressive to whether it can fit real delivery operations. Rain, snow, mud, darkness, and damaged paths create failures that clean demonstrations hide. Testing those conditions now helps Amazon find weak points before it puts more robots beside drivers and customers.

Q6So what should we watch next?

Watch the fleet size, human supervision, failed deliveries, operating cost, and how often a person must rescue the robot. The key step is not another video of one machine climbing stairs. It is dozens or hundreds of robots completing deliveries every day in different cities and seasons. If RIVR can do that, Amazon may finally have a practical robot for the final 100 yards.