Autonomous Robot Deliveries

TOBY SNUGGS We operate self-driving delivery robots on the Uber Eats platform. We have the robots stationed in various areas around the operating area, waiting for orders to be placed. And this is one of the restaurants that we pick up food from and deliver to Uber Eats customers.

If you live in this area, and the restaurant you want to order food from is part of the robot program, the Uber Eats app would automatically assign a robot to the order to be delivered to your location. It drives on sidewalks and crosses roads just like a pedestrian. You will actually see the robot traveling on a map. The robot doesn’t go inside buildings, so the customer does need to come outside and open the hatch of the robot through the app.

Our company actually originates from level four self-driving car technology. Our tech teams repurpose that self-driving car technology into a sidewalk delivery robot using the same software on the back end and also, most importantly, the same sensor sets.

So, the lidar is one of the most important pieces of technology on our self-driving cars and robots. It builds a 360-degree picture around the robot. In addition to that, we have six cameras around the robot. And then we do have ultrasonic sensors, detecting anything that comes close to the robot. If something like a child falls over in front of the robot, the robot comes to a complete stop. So the data from all of those sensors feeds back into our machine learning, AI powered software. And that’s really the brains that pulls all this data together and enables the robot to make decisions about the things that it’s seeing, perceive what’s happening, and then lets the robot continue on its journey.

The critical thing for self-driving technology is just the amount of data that gets fed into our machine learning software. So the more miles that our robots drive, the more it understands potential scenarios it will come across. Busy intersections, it will gradually learn which of the ones to avoid, how the traffic signals work in the particular city that it's in. What one robot is picking up and perceiving is being fed back into the software that other robots can then use as well.

When we move into a new operating area, the first thing we do is we map the area. We typically use one robot to do this. We currently have 30 robots here. We can add many, many, many, more robots. They all utilize the same map. When an incident does happen, like a tree falling across the sidewalk, and the robot comes to a stop, the robot actually alerts the remote operators that can help the robot make a decision to navigate around that situation. So even though we have remote operators always monitoring our robots, our robots actually operate over 99 percent autonomously.

So, I personally can see the future, these robots become much more commonplace. Robots are not designed to replace humans, they’re there to really augment human delivery. And right now we’re really seeing a situation where there aren’t enough couriers around, demand is increasing. So, a robot can fulfill some part of that delivery chain.