[ad_1]
Hello, robot!
Author: Ahti Heinla, Co-founder and CTO of Starship Technologies
I see robots every day. I saw them sliding down the sidewalk at pedestrian speed and stopped to make sure it was safe to cross the road. Sometimes I even see them talking with pedestrians. This is a glimpse of a technically minded fantasy-an artificial intelligence wonderland. But this is not an illusion, nor a dream. This is the reality that our dedicated visionary team has established over the past 5 years; we bring the future into the present.
Just a few years ago, these robots needed a little human support and accompanied their journey, just like the format followed by self-driving car manufacturers, they used “safe drivers” to test their cars in public.
About 18 months ago, Starship became the first robot team to start regular operations in public places, without the use of safety drivers; we let our robots explore the world on their own. Now, we run our robot network in multiple cities around the world every day, delivering dinner, parcels and groceries for people.
Knowledge sharing is the knowledge gained
It’s exciting to be the first.
When I was the founding engineer of Skype, we were the first company to make voice over IP accessible in a practical way; we are now trying to do the same with robots in public places. Over the past four years, our engineering team has worked behind closed doors and has achieved major breakthroughs and amazing experiences.
I want to share with you some details of our technological journey. In the coming weeks and months, other members of the Starship engineering team will also share various aspects of their journey.
In the process, we have studied computer vision, path planning and obstacle detection-these topics have been in-depth research in the field of academic robotics. In fact, Starship was originally a research project, but quickly transformed into a practical and practical delivery operation.
This means that in addition to fine-tuning the Levenberg-Marquardt algorithm for nonlinear optimization, we must also develop software to:
- Automatically calibrate most of our sensors-after all, we don’t want to spend hours manually calibrating them; we have built hundreds of robots and are currently preparing for larger operations.
- Predict how much energy will be drawn from the robot’s battery for each trip-so we can schedule which robot to send based on their battery status.
- Predict how many minutes it will take for the restaurant to prepare food-so the robot will appear in time!
Most autonomous robots that exist in the world today are expensive. They are manufactured as technology demonstration or research vehicles and are not used for commercial operations. The cost of a sensor package for autonomous devices alone can be as high as $10,000. This simply does not work in the field of delivery, it is not a luxury industry that can charge premium prices.
The trunk of autonomous driving research vehicles often has 3 kilowatts of computing power; it is impractical for small, safe delivery robots. Therefore, part of our engineering journey is to design for lower unit economics. Here are some topics we must consider:
- Advanced image processing on low-end computing platforms.
- Solve hardware problems in the software.
- Track how often and why the robot needs maintenance.
- Develop an advanced route planning system to ensure that we use our robot network effectively.
Before we made the first plastic body of the robot, it was also a long visual design journey involving hundreds of sketches, drawings, and surveys.
Back in the early days when we were still in stealth mode, we didn’t want to reveal the appearance of our robot. Regular public testing requires creative use of garbage bags, taped to the body of the robot as a disguise!
Constructing practical robotics is a fusion of science, systems engineering and hacker technology. This mixing of different disciplines is the core feature of Starship. Nothing in robotics is simple.All your knowledge of the situation is probabilistic; all sensors have failure modes and failures, even seemingly simple tasks, such as Stop the robot at the obstacle Can be your own small research project.
Starship is a fast-growing start-up, and it is important not to just become a large research project. Engineers who are excited about Starship are often not pure scientists, not pure hackers, or pure engineers; they have several such characteristics that can be used according to the task at hand. We need to quickly implement complex technical solutions within the resource constraints of low-cost hardware.
Ingenuity and resourcefulness are precious skills.
A long week in Starship
At the beginning of this week, our team will implement a new algorithm to detect curbs in the point cloud and backtest it against the complete test case database overnight. They will test it in our private test before the end of the year It will be tested on site for one week.
It will appear on the street next Monday, and the team has reported their progress at our engineering meeting on Monday. On most Mondays, some members of the engineering team reported that in the previous week, at least one indicator had achieved a gain of more than 300%.
Data is the result and facilitator of scale
Indicators and data have become an important part of the Starship project.
You see, when we just started, we had no data-we hadn’t driven many cars yet. Every day we modify our robot (yes, just the one at that time), take it to the sidewalk, and see how it performs. We have a lot of them now, and they drive every day—engineers can’t directly observe too much.
Thanks to this data, we can now see the performance of our robots, hundreds of them. We can organize weekly “data diving” seminars where engineers can share survey results and watch random deliveries to understand what they are doing.
As we work to make our robots drive more smoothly, we analyze the data in the “acceleration events” table of the data warehouse; there are at least 1 billion rows in this table. Other tables include “road crossing incidents”, our maps, each command that each robot receives from our server, and the data collected from each delivery they perform.
Four years ago, we did not have these. Back when we were just starting out-without commercial delivery yet-I often had to convince people that robot delivery was indeed effective. People found it hard to believe and quickly pointed out various reasons.
Do doubts and fears always accompany new technologies?
A few years ago, I landed at JFK Airport in New York with a robot. The customs officer obviously asked, “What is this?” I explained that it was a sidewalk delivery robot, and he replied, “Man, this is New York! It was stolen in minutes!”
In fact, almost everyone at the time thought these robots would be stolen-I’m sure they could be stolen (postal delivery vans were stolen, even if very rarely). So far, our robot has traveled more than 200,000 kilometers (130,000 miles), and we have not seen this problem.
Of course there are safety features. The robot has a siren and 10 cameras. It is constantly connected to the Internet and knows its precise position with an accuracy of 2 cm (thanks to the Levenberg-Marquardt algorithm mentioned above, and 66,000 lines of automatically generated C++ code, which allows us Of robots can use it).
People also think that pedestrians may be afraid of robots on the sidewalk, or may not accept their presence. Will people call the police? Honestly, we are not sure about this either! However, once we put one of the robots on the sidewalk, we were taken aback.
What happened next surprised us: people just ignored it. The vast majority of the public does not pay any attention to robots, and people are certainly not afraid of them even if they are the first to see it. Others will take out their phones and post their thoughts on the future on Instagram.
This is what we want.
We want people to pay attention to our robots as much as they do to dishwashers. This mode of silently accepting robots, as if they have been by our side, repeats itself in every city around the world we have operated.
It got better. Once people understand that these robots provide useful services to the community, they will develop an affinity with them. Children will even write to thank the robots, we have a “thank you letter wall” to prove this!
Automating last mile delivery is no easy task, and we know it will be a bold project. We have always known that there is more than one basic obstacle to be resolved—the result is hundreds of obstacles! But we realized long ago that all these problems are solvable-they only require ingenuity and perseverance.
Some start-ups are like sprints at first, piecing together a minimum viable product in 3 months. For Starship, it is more like a marathon-it requires a lot of continuous effort, but the end result will bring huge benefits to the world.
Last mile delivery is one of the world’s industries that have seen almost no technological disruption since the adoption of cars. The Starship team is looking to change this situation. We have completed more than 20,000 deliveries and our work is progressing smoothly.
If you are interested in learning more, check out our second engineering blog post on neural networks and how they power our robots here — https://medium.com/starshiptechnologies/how-neural-networks-power-robots-at-starship-3262cd317ec0
[ad_2]
Source link
