Amazon is expanding its work on Amazon Scout, the company’s land-based delivery robot. The delivery giant is hiring a team of engineers and developing new software to help simulate real-world conditions its robots will face out on the streets.
With these developments, it seems like the delivery bot Amazon first began testing in 2019 is now showing some real promise. According to Amazon, its new team will consist of more than 24 engineers based in Helsinki. Their chief goal will be to “develop 3D software to simulate the complexity of real life and ensure Scout can safely navigate around obstacles while making deliveries.”
That’s quite vague, but there are some open positions tagged Scout at Amazon that give some clue into the nature of the work. The company is hiring for software engineering roles and machine learning leaders. That paired with the early 2021 acquisition of 3D modeling company Umbra suggests Amazon hopes to speed up Scout’s wisdom with simulated experiences.
That kind of focus is common when training robots, and keeps training in the safety of a 3D world. Instead of learning one-off situations out in the real world with someone’s new headphones or plant pot, Scout can figure out situations in a digital world of no consequences.
Amazon said the Helsinki office would grow over time and that this represents work at “the forefront of robotics and autonomous systems technology.”
Compared to Starship technologies or Kiwibot, Amazon is the new bot on the block. In the first-quarter filing from Amazon, the company reported that it had “delivered tens of thousands of packages” across California, Georgia, Tennessee and Washington. Starship, however, just reported 1 million deliveries and Kiwibot announced 150,000 deliveries with its adorable bot.
That puts Amazon quite a bit back from the pole position in robotic delivery, but given the investment here and the obvious vertical integration with the rest of Amazon, the other players are probably getting a little anxious even if Amazon hasn’t made a big push into prepared food.