Hello, Robot !. I see robots every day. I saw them… | by Ahti Heinla | Starship Technologies
By Ahti Heinla, co-founder and CTO of Starship Technologies
I see robots every day. I saw them running down the sidewalks at pedestrian speed, stopping to make sure it was safe to cross the road. Sometimes I catch them talking to pedestrians. It’s a glimpse into the fantasies of the technology -minded – a Wonderland of AI. But it’s not holiness, it’s not a dream, it’s a reality built by our team of dedicated visions over the past 5 years; we bring the future to the present.
Just a few years ago, these robots needed a little bit of human support and were accompanied on their journeys, as followed by a format that followed autonomous cars, testing their cars in public. using ‘safety drivers’.
Starship became the first robotics team to start operating regularly in public spaces about 18 months ago, without using safety drivers; We allow our robots to explore the world on their own. We now run our network of robots daily in many cities around the world, bringing people to their dinners, parcels and groceries.
Shared knowledge is acquired knowledge
Exciting to go first.
Back when I was a founder engineer at Skype, we were the first to access Voice over IP in a practical way; we are working now to do the same with robots in public space. For four years, our engineering teams have been working behind closed doors on what has been a significant achievement and surprising experience.
I would like to share some of the details from our technical journey with you. In the coming weeks and months, other members of the Starship engineering team will also share aspects of their journey.
With this journey we worked with computer vision, route planning and obstacle exploration – topics well explored within the realm of academic robotics. Actually, Starship started out as a research project, but was later moved into a moving practical delivery operation.
This means that in addition to the good alignment of the Levenberg-Marquardt algorithm for nonlinear optimization, we need to develop the software to:
- Automatic calibration of most of our sensors – after all, we don’t want to spend hours calibrating them by hand; we have built hundreds of robots and are now preparing for a much larger scale operation.
- Predict how much power will be drawn on each trip from a robot’s battery – so we can orchestrate which robot will be sent, adapted to the state of the battery.
- Predict how many minutes it takes for a restaurant to prepare food – so the robot shows up at the right time!
Most of the autonomous robots in the world today are expensive, they are built as technology demonstrators or research vehicles and are not used for commercial operation. A single sensor package for an independent device can cost up to $ 10,000. Not only does it operate in the delivery space, it’s not a luxury industry where you can pay a premium.
Self-contained car controllers typically have 3 kilowatts of main computing power; impractical for a small, safe delivery robot. As such, part of our engineering journey is about designing for a much smaller unit economy. Here are some topics we need to consider:
- Advanced image processing on a low end computational platform.
- Working around software hardware issues.
- Keep track of when robots often need to keep up, and why.
- Develop advanced route planning systems, to ensure we make good use of our network of robots.
It was a visual design trip as well, involving hundreds of sketches, drawings and surveys before we made the first plastic body of our robot.
Back in the early days when we were still in stealth mode, we didn’t want to reveal what our robots looked like. The public should regularly test the creative use of a trash can, which is attached to the body of the robot as a cover!
Building practical robotics is a mixture of science, systematic engineering and hacking. This mix of different disciplines is the main character of Starship. There is nothing simple about robots. Everything you know about the situation is possible; all sensors have failure and error modes, and even such a simple task as make the robot stop at obstacles be your own small research project.
Starship is a fast -paced business start -up and essentially not going to be a big research project. The engineers who are excited about Starship are always not pure scientists, not pure hackers, not pure engineers; they have a number of characteristics and can be used as appropriate to the task being prepared. We need complex technical solutions to be implemented immediately and within the constraints of hardware resource that are low cost.
Ingenuity and creativity are essential skills.
A week long hours in Starship
At the start of the week our team will implement a new algorithm to detect curbs from the point cloud and test it back against an entire test case overnight, they will test it live at our private test-ground at the end of the week.
It will take to the streets next Monday, with the group already reporting on their progress during our Monday’s Engineering Meeting. Most of the Mondays some of the engineering team reported a 300% + gain on any one scale achieved, just last week.
Data as outcome and measure manager
Basics and data have become a big part of Starship’s engineering.
You see, back in the day we had no data – we hadn’t driven much. Every day we updated our robot (yep, the only one before), took it to the sidewalks and saw how it was made. There are a lot of us now, driving every day – which is a lot for engineers to directly observe.
Thanks to the data, we can now see how our robots are made, hundreds of them. We can organize weekly ‘data dive’ seminars, where engineers share insights and watch random deliveries to keep in touch with their activity.
If we’re working on making our robots drive more smoothly, we’re analyzing the data in the “acceleration activities” table in our Data Warehouse; there are at least 1 billion rows on that table. The other tables include ‘road crossing events’, our maps, every command received by each robot from our servers, and clearly collected data from every delivery they make.
Four years ago, we didn’t have it. Back when we were just starting out – and before running commercial deliveries – I always convinced people that robotic delivery would really work. People have a hard time believing and are quick to point out various reasons why.
Doubt and fear always accompany new technology?
Many years ago, I arrived at JFK airport in New York carrying a robot in my luggage. The man clearly asked the customs: “What is this thing?” I explained that it was a sidewalk carrying robot, to which he replied: “Friend, this is New York! It will steal in a few minutes! ”
In fact, in the past almost everyone thought these robots would steal – I’m sure they probably did (postal delivery vans were stolen, albeit rarely). To date our robots have been driving over 200,000km (130,000 miles) and we have yet to see that problem.
There are security side courses in place. The robot has a siren and 10 cameras, it is always connected to the Internet and knows its exact location with an accuracy of 2cm (thanks to the above-mentioned Levenberg-Marquardt algorithm, and 66,000 lines are automatically generated C ++ code that our robots can use. to use it).
People also think that pedestrians may be afraid of robots on the sidewalk or may not accept their presence. Will the police call people? Honestly, we’re not sure about it! However, once we put one of the robots on the sidewalk, we were shocked.
We were surprised by what happened next: people just ignored it. The vast majority didn’t pay any attention to the robots, even those who saw them for the first time, and people were definitely not scared. Some will pull up their phones and post on Instagram about how they see the future.
And that’s what we want.
We want people to pay as much attention to our robots as they do to their washing machines. This pattern of quietly accepting robots as if they were always with us repeats itself in every town around the world we work with.
Healed. Once people learn that these robots provide a useful service to the neighborhood, they develop a relationship with them. Kids are still writing thank you letters to robots, we have a ‘wall thank you letters’ to prove it!
Automating last-mile delivery is never easy, and we know it’s an attractive project. We also all know that there is more than one basic roadblock that needs to be addressed – there are hundreds of roadblocks! But we know before that all problems can be solved – they just need ingenuity and perseverance.
Some startups start out like running a sprint, throwing in a Minimum Viable Product in 3 months. For Starship it’s more like a marathon – it takes a lot of perseverance, but the end result brings a lot of benefits to the world.
Delivery of the last mile is an industry in the world that has seen little destruction in technology since the car was adopted. The Starship team is looking to change that, and with over 20,000 deliveries under our belt, we’re almost done.
If you’re interested in finding out more, check out our second blog post on Neural Networks Engineering and how they power our robots here- https://medium.com/starshiptechnologies/how-neural-networks-power-robots-at-starship-3262cd317ec0