By Ahati Heinla, co-founder and CTO of Starship Technologies
I see robots every day. I see them gliding along the sidewalk at pedestrian speeds, stopping to make sure the road is safe to cross. Sometimes I even catch them talking to pedestrians. It’s a glimpse into the imagination of technically minded people – an AI wonderland But it’s not a hallucination, it’s not a dream, it’s a reality that our team of dedicated dreamers has created over the last 5 years; We have now come to the future.
Until a few years ago, these robots needed some humanitarian assistance and had a companion on their journey, a format that many autonomous car manufacturers follow, who test their cars in public using ‘safety drivers’.
Starship is the first robotics team to start regular work in public space about 18 months ago, without the use of a security driver; We let our robots explore the world on their own. We now operate our robot network every day in different cities of the world, bringing people their dinner, parcels and groceries.
Sharing knowledge is the acquired knowledge
It’s exciting to be first.
When I was a founding engineer on Skype, we were the first to make Voice over IP practically accessible; We are now working to do the same thing with robots in public space Over four years, our engineering teams have worked behind closed doors, which has been a remarkable breakthrough and an amazing experience.
I would like to share with you some details of our technological journey. In the coming weeks and months, other members of the Starship Engineering team will share their journey directions.
In this journey we have worked on computer vision, path planning and obstacle identification – topics that are well researched in the field of academic robotics. In fact, Starship began as a research project, but was soon transformed into a functional, practical delivery operation.
This means that in addition to fine-tuning the Levenberg-Marquard algorithm for non-linear optimization, we need to create software:
- Most of our sensors calibrate automatically – after all, we don’t want to spend hours calibrating them; We have built hundreds of robots and are currently preparing for a large-scale operation.
- Estimate how much power will come from a robot’s battery on each trip – so we can determine which robot to send based on the condition of the battery.
- Estimate how many minutes it takes to prepare a restaurant meal – so the robot will show up just in time!
Most of the autonomous robots that exist in the world today are expensive, they are made as technology demonstrators or research vehicles and are not used for commercial activities. A single sensor package for an autonomous device can cost upwards of $ 10,000. It will not only work in the delivery space, it is not a luxury industry where you can charge a premium.
Autonomous driving research vehicles often have 3 kilowatts of computing power in the trunk; Unreal for a small, secure delivery robot. Therefore, part of our engineering journey is designed for lower unit economy. Here are some things to consider:
- Advanced image processing on a low end computational platform.
- Dealing with hardware issues in software.
- Tracking how often robot maintenance is required, and why.
- Creating advanced route planning systems to ensure that we use our robotic network efficiently.
It has also been quite a journey in terms of visual design including hundreds of sketches, drawings and surveys before we created the first plastic body of our robot.
In the early days when we were still in stealth mode, we didn’t want to reveal what our robots looked like. Regular public testing requires creative use of a trash bag, taped on the robot’s body as a camouflage!
Creating practical robotics is a mix of science, methodological engineering and hackery. This mix of different disciplines is the main feature of Starship. Nothing is easier in robotics. All your knowledge of the situation is possible; All sensors have failure modes and errors, and even a seemingly simple task such Interrupts the robot Could become his own small research project.
Starship is a fast-moving startup business and it’s important not to just turn it into a big research project. Engineers excited about the starship are often not pure scientists, not pure hackers, not pure engineers; They have a number of these features and can be used as a handy task. We need complex technical solutions for rapid implementation within the constraints of low cost hardware resources.
Cleverness and wealth are valuable skills.
A week long time at Starship
At the beginning of the week our team will implement a new algorithm to detect the curb from Point Cloud and re-test it against an entire test case database overnight, they will test it live at our personal test-site by the end. Week
It will be on the road next Monday, the team will already report on their progress during our Monday engineering meeting. On most Mondays, some members of the engineering team are reporting a 300% + profit on at least one of the metrics acquired just a week ago.
Data as a result of scale and benefits
Metrics and data have become a big part of starship engineering.
You see, we didn’t have any data when we were just starting out – we haven’t driven much yet. Every day we would change our robot (yes, just one then), take it to the sidewalk and see how it works. We now have a lot, a lot of autonomous driving every day – much more for the engineers to observe directly.
Thanks to the information, we can now see how our robots work, hundreds of them. We can organize weekly ‘Data Dive’ seminars, where engineers share results and see random deliveries to keep in touch with their work.
While we’re working to make our robots drive more smoothly, we analyze data from our data warehouse’s ‘acceleration event’ table; There are at least 1 billion rows on that table. Other tables include ‘Road Crossing Events’, our maps, every command each robot receives from our servers, and obviously data collected from each of their deliveries.
Four years ago, we had none of this. When we were just starting out – and still not running commercial deliveries – I often had to convince people that robotic delivery really works. People were hard to believe and they were quick to point out various reasons.
Do doubts and fears always accompany new technology?
Several years ago, I landed at JFK Airport in New York with a robot in my luggage. The customs man asked explicitly: “What is this?” I explained that it was a sidewalk delivery robot, to which he replied: “Dude, this is New York! It’ll be stolen in a few minutes!”
In fact, at the time, almost everyone thought these robots would be stolen – I’m sure they would (the postal delivery van was stolen, though very rarely). To date, our robots have traveled more than 200,000 km (130,000 miles) and we have yet to see that problem.
There are, of course, security features. The robot has a siren and 10 cameras, is constantly connected to the Internet and knows its precise position with 2cm accuracy (thanks to the above-mentioned Levenberg-Marquard algorithm and 66,000 lines of automatically generated C ++ code that enable our robot to use). .
People also thought that pedestrians might be afraid of robots on the sidewalk or not accept their presence. Will people call the police? Honestly, we weren’t sure about that! However, once we put a robot on the sidewalk there, we were quite surprised.
What happened next surprised us: people ignored it. The vast majority of the public paid no attention to the robots, even those who saw it for the first time, and people certainly did not fear. Others will take out their phones and post on Instagram about how they see the future.
And that’s what we wanted.
We want people to pay as much attention to our robots as they do to their dishwashers. This pattern of silent acceptance of robots, as if they were always with us, has been repeated in every city in the world where we have worked.
It’s getting better. When people find out that these robots provide a useful service in the neighborhood, they form a friendship with them. Kids even write letters of thanks to the robot, we have a ‘thank you letter’ to prove it!
Automating last-mile delivery will never be easy, and we knew it would be a daunting task. We also knew that there would be multiple basic roadblocks that needed to be addressed – there were hundreds of obstacles! But we realized long ago that all these problems are solvable – they just need ingenuity and perseverance.
Some startups start out running sprints, throwing together a minimum effective product in 3 months. It’s like a marathon for Starship – it takes a lot of effort, but the end result brings huge benefits to the world.
Last Mile Delivery is an industry in the world that has seen little technical disruption since the adoption of the automobile. Starship’s team is on a quest to change that, and with over 20,000 deliveries under our belt, we’re well on our way.
If you are interested in learning more, check out our second engineering blog post on Neural Network and how they power our robots here – https://medium.com/starshiptechnologies/how-neural-networks-power-robots-at-starship -3262cd317ec0