How Starship Delivery Robots know where they are going | by Joan Lääne | Starship Technologies
(plus how you can make your own 1: 8 papercraft robot model)
by: Joan Lääne, Mapping Specialist, Starship Technologies
Every September when the new school year begins, many first-graders are a little scared of the unknown. Not just about starting school and the new people they meet, but also about the journey they have to make every day. They need to know and remember how to navigate the world and the way to and from their classroom on their own. It may be easiest for a parent to accompany their child on the first trips back and forth to make them more familiar with the trail, often pointing to some interesting landmarks along the way such as, tall or bright buildings. or traffic signs. Later it may not be important for the child to go to school and remember the method. The child makes a mental map of the world and how to navigate it.
Starship Technologies provides a convenient last-mile delivery service with hordes of sidewalk delivery robots navigating the world every day. Our robots have completed over 100,000 deliveries. To get from point A to point B the robots have to plot a route forward which instead requires a map class. Although many publicly available mapping systems such as Google Maps and OpenStreetMaps, they have the limitation that they are designed with car navigation in mind and mostly focus on mapping car routes. Because the robots that deliver it travel the sidewalks, they need an accurate map of where to safely travel on the sidewalks and where to cross the streets like a child needs a mental map of how to get there. at school safely and on time every day. So how is this map made?
The first step in creating a map for delivery robots is to examine the area of interest and create a preliminary map (2D map) over the satellite image in the form of simple interconnected lines representing on sidewalks (green), crossings (red), and walkways (purple) as shown in the image below.
The system treats this map as a node graph and it can be used to create a path from point A to point B. The system can determine the shortest and safest path to the robot and calculate it. also the distance and time it will take to drive this route. The advantage of this process is that everything can be done remotely before any robots reach the site.
The next step involves showing the robots what the world looks like. Similar to the parent -child similarity, robots require little hand grasping the first time exploring an area. When the robot first drives, the robot’s cameras and many sensors collect data about the world around it. This includes thousands of lines that range from identifying edges of various shapes, such as for example buildings, street light poles and rooftops. The server can then offline a 3D world map from the lines that can be used by the robot. Like the child, the robot now has a model world with guidance posts and can understand where it is at any time.
Since our robots need to cover different areas at the same time to complete all deliveries, in order to be efficient the different maps need to come together to create a unified 3D map of the given area. The combined map is made per piece by processing different pieces in the new area until finally the map looks like a multi-finished jigsaw puzzle. The server will place this map based on the line data collected by the robot previously. For example, if the same rooftop is found by two robots, then the software knows how to connect it to the rest of the map. Each color line in the image below represents a mapping trip piece added to the map.
The final step in the mapping process, before the robots are able to drive completely independently, is to calculate exactly where and how wide the sidewalk is. This is done by processing camera images recorded by the robot while exploring the area as a reference as well as mixing a previously created 2D map based on the satellite image.
During this process many details are added to the map to accurately pinpoint the safe areas where the robots can drive.
Of course, the world around us is not static. There are daily and seasonal changes in the landscape, constructions and repairs, which are changing the look of the world. How does this affect area maps for robots? In fact, the robot software handles small to medium area map changes very well. 3D models are so powerful and filled with a lot of data, that a tree cut down here or a building destroyed there often doesn’t pose a challenge to the robot’s ability to position it or use the map. And, in addition as the robot drives each day it continues to collect a lot of data that is used to update the 3D maps over time. But if an area is completely renovated, or new sidewalks are being built, then the solution is quick. The map must be updated using new data gathered by a robot. Afterwards, other robots can also drive independently in the same area as if nothing had happened. Maintaining maps is paramount to keep robots from driving safely and independently.
As you will not hesitate to say by now, I enjoy playing with concepts in 3 dimensional space. Ever since I played the first 3D first person shooter computer game (Wolfenstein 3D), the 3D world in the digital domain has become my interest. I wanted to create my own 3D worlds for computer games, so I found ways to edit the existing levels of the game. Later, I also tried my hand at 3D computer modeling, which I found to be interesting. With the popularity and accessibility of 3D printers, I also started with physical printing models. But before that, during the summer at school, I wanted to make papercraft models in different buildings and cars. It’s a quick and inexpensive way to make something with my own hands, although it’s also interesting to see how a 2D layout works on a piece of paper, with little cutting, folding and glue, become a 3D model. Basically, making papercraft into a 3D object or “opening”, in a sense, is the reverse of mapping. It creates a 2D layout of the face of a 3D object.
Since I have a love for papercraft I decided to make one for our Starship delivery robots. The purpose was to create this model so that others could enjoy the same passions I created to create their own version of our delivery robots. Making a paper model is a fun challenge, and once it’s done it’s also made for a beautiful decorative item as well. Just like making 3D maps for a robot, creating a papercraft model requires precision, accuracy, and spatial thinking about how all the parts come together. Good patience too.
I made some instructions for you so you can make your own robot that carries papercraft and I want to see your efforts. Have fun and good luck creating your own role model of the delivery robot!
Please post a photo of your robot on Instagram and tag @StarshipRobots so I can find them!
Please find the model of the robot delivery robot to send Starship and instructions HERE
© Starship Technologies. The design of the Starship® delivery robot and aspects of the technologies are described as proprietary and protected by copyright and other intellectual property laws