Author Ahti Heinla, co-founder of CTO at Starship Technologies
I see robots every day. I can see them running in the street, stopping to make sure it is safe to cross the street. Sometimes I find them talking to pedestrians. It is a reflection of the mindset of professional minds – the amazing world of AI. But this is not true, there are no dreams, it is a reality that our volunteer vision team has made over the last 5 years; we have brought it in the future.
More recently a few years ago, these robots need human help and accompany them on their journeys, as do self-propelled car manufacturers, who test their vehicles openly using ‘safe drivers’.
Starship became the first group of robotics to start operating more frequently in public places about 18 months ago, without the use of safe drivers; we allow our robots to explore the world on their own. Now we use our robots every day in several cities around the world, bringing people dinner, packages and food.
Sharing information and acquired knowledge
It’s a pleasure to be the first one.
Back when I was a startup engineer in Skype, we were the first to make Voice over IP available in a practical way; now we are working to do the same with robots in public places. For four years, our engineering teams have been working without doors that have been running smoothly and surprisingly well.
I want to share with you more about our professional journey. In the coming weeks and months, some members of the Starship engineering team will share some of their tours.
Along the way we have worked with computer vision, design solutions and detection barriers – studies that are well researched in the field of educational robotics. True, Starship started out as a research project, but soon evolved into a more efficient, effective delivery system.
This means that in addition to fine-tuning the Levenberg-Marquardt algorithm to achieve optimal performance, we have developed programs that:
- Just add a few of our sensors – after all, we don’t want to waste many hours and test them manually; we have made hundreds of robots and here we are preparing for a big project.
- Note the amount of energy that each trip will receive from the robot battery – so that we can configure the robot to send, depending on the battery status.
- Be aware that it may take several minutes for the restaurant to prepare the food – for the robot to appear on time!
Many of the independent robots that exist in the world today are expensive, built as technology or search engines and not used for commercial purposes. A single sensor package for an independent device can cost up to $ 10,000. This will not work in the delivery area, it is not a high-end business where you can pay a lot of money.
Self-propelled research vehicles usually have 3 kilowatts of computing power in total; useless for a small, secure delivery robot. As a result, part of our engineering journey has been about creating a small economy. Here are the topics to consider:
- High-resolution image editing on a low-reading platform.
- Working around hardware problems in software.
- Investigate how robots need to be repaired, and why.
- Developing high-throughput systems, ensuring that we use our robots efficiently.
It was also a spectacular journey, combining hundreds of photographs, drawings and explorations before making the first plastic body of our robot.
Back in the early days when we were still on the road, we didn’t want to reveal what our robots looked like. Frequent public testing involves the wise use of the trash can, which is printed on the body of the robot as a hidden object!
The development of effective robots is a combination of science, systematic engineering and destruction. This integration of various systems is a Starship tradition. Nothing is easier in robotics. All your knowledge of the situation is possible; All sensors have failure modes and errors, and even seemingly simple tasks like making the robot to stop at obstacles it could be his little research work.
Starship is a fast-growing start-up business and it is important that it does not become a major research activity. Engineers interested in Starship are often not perfect scientists, not destroyers, not perfect engineers; they have a number of these characteristics and can use them as appropriate to the task at hand. We need sophisticated technical methods to be implemented quickly and within the hassles of cheap tools.
Skill and skill are invaluable.
The Sabbath is a long time in Starship
Earlier in the week our team will use a new way to detect cloud-based screens and test them on a complete overnight trial database, which will be tested live on our secret testing site by the end of the week. sabbath.
It will be on the streets next Monday, the team is already talking about how it is going at our Engineering Meeting Monday. On Monday most of the team of engineers report that they have achieved 300% + at the same level achieved, last week.
Data as a follow-up by the scale manager
Metrics and data have become an integral part of Starship engineering.
You see, back when we were just starting out we had no data – we hadn’t traveled much yet. Every day we would change our robot (yes, the only one), take it to the streets and see how it went. Now we have many, traveling around every day – too many for the engineers to see directly.
Thanks to the data, we can now see how our robots perform, hundreds of them. We can organize weekly ‘data dive’ seminars, where engineers share their findings and view randomly posted to connect with their work.
As we strive to keep our robots running smoothly, we analyze the content of our ‘Data Warehouse’ table; there are at least 1 billion lines on the table. Some of the tables include ‘cross-road events’, our maps, every command any robot has received from our servers, and obviously the data that is collected from every movement they perform.
Four years ago, we didn’t have this. Back when we were just starting out – and we weren’t running a business – I often had to convince people that robot shipping really worked. People found it difficult to believe and were quick to point out various reasons.
Do doubts and fears always accompany new technology?
A few years ago, I arrived at JFK Airport in New York with a robot in my backpack. “What is this?” The boy asked. I explained that it was a roadblock, and he replied: “Dude, this is New York! They’ll be stolen in a few minutes! “
In fact, at that time almost everyone thought these robots would be stolen – I believe they would probably be stolen (postal cars are stolen, albeit often). So far our robots have traveled over 200,000km (130,000 miles) and we have not seen a problem.
There are of course safety features in place. The robot has a siren and 10 cameras, is constantly connected to the internet and knows its exact location with an accuracy of 2cm (thanks to the aforementioned Levenberg-Marquardt algorithm, and 66,000 self-propelled C ++ lines that support our robots. work).
People also thought that pedestrians might be intimidated by traffic lights or that they would not acknowledge their presence. Will people call the police? In fact, we could not even be sure of that! However, as soon as we put one robot on the side of the road, we were surprised.
What happened next shocked us: people just ignored it. Most people did not pay attention to the robots, even those who saw them for the first time, and the people were certainly not intimidated. Some are able to pull out their phones and post on Instagram of how they view the future.
And that is what we wanted.
We want people to take care of our robots just as much as they do with their dishwashers. The practice of quietly accepting robots as if they were always with us has been repeated in every city in the world in which we live.
It feels good. When people know that these robots are useful to their neighbors, they are more likely to associate with them. Kids even write thank-you letters for robots, we have a ‘wall of thank-you letters’ for sure!
Sending final miles was not easy, and we knew it would be a daunting task. We also knew from the beginning that there would be more than one road obstacle that needed to be overcome. But we have long known that all these problems are possible – they just need wisdom and perseverance.
Other triggers start as a run, throwing together the Less Possible Thing in three months. For Starship it is like a race – constant effort is required, but the result brings the best to the world.
Sending miles is one of the few industries in the world that has experienced a slight technological breakdown since the introduction of automobiles. The Starship team wants to change this, and with more than 20,000 shipments under our belt, we are on our way.
For more information, see the second page of our Engineering blog on Neural Networks and how they make our machines here – https://medium.com/starshiptechnologies/how-neural-networks-power-robots-at-starship-3262cd317ec0