Autonomy Explained by the Experts at Kiwibot

June 29, 2022

A robot rolling along a sidewalk, delivering food, makes us feel closer to the future.  Kiwibot does it every day, greeting people and bringing you the food you love. Sci-fi? Is it possible for them to do it all independently, without any human interaction? To answer these and other questions, meet the main authors of Autonomy Development at Kiwibot, the ones who turn fiction into reality.

John Betancourt

I

ntroducing John Betancourt, head of AI & Robotics, with a Bachelor's Degree in Electronic Engineering, and expertise in software, artificial intelligence, computer vision, and of course, robotics. Just like the conductor of an orchestra, he leads the team’s engineers toward resolving the fundamental problem of autonomy, while they bring solutions and make the magic happen: Kiwibots strolling along streets. More on this a little further down.

Carlos Alvarez


Then there’s Carlos Alvarez, our Senior Machine Learning Engineer and expert in image processing. Although it may sound straightforward, this bit is where science really comes to the fore. In order for Kiwibots to resemble EVA, WALL-E’s sleek companion, Carlos spearheads   the training of each Kiwibot brain to automatically recognize which part of the image corresponds to things like a sidewalk, streets, people, traffic lights, etc.

Consequently, this magic is only possible after an intensive on-site data collecting, processing, and labeling phase. After that, with all this information, our robot builds a model of the environment, which is used to make decisions about where it can or cannot go.

Both John and Carlos lead the mission to crack the still-unfinished problem of the robotics world: Full Autonomy. And yes, for everyone working on this, it remains unsolved. Let’s explain a bit more…

What is autonomy?

Simply put, in robotic terms, autonomy is being able to perform a task without any human intervention.

No alt text provided for this image


Let’s start at the beginning. Robots seek to imitate the behavior of any system that moves or has locomotion, such as arms, cars, animal imitations, etc. They have different purposes and fields of action, like manufacturing, supply chain, entertainment, medical support; or in our case, deliveries.

In technical terms, the Society of Automotive Engineers (SAE),commonly referenced as the industry’s most-cited source for driving automation, talks about “[motor] vehicle driving automation systems that perform part or all of the dynamic driving task (DDT) on a sustained basis.” This definition is broad and can be adapted to any robotic system whether it is aerial, land or marine.

According to MIT’s book ‘Autonomous Robots, From Biological Inspiration to Implementation and Control,’ “autonomous robots are intelligent machines capable of performing tasks in the world by themselves, without explicit human control.”

Now, let's imagine this: you are walking on a street and you see a robot crossing it. There is some obstacle in the middle, a garbage can, and further, someone is pushing a stroller. Now the robot needs to evaluate the risks, decide to go slower or stop, then avoid the obstacles and say “Hi” to the woman all by itself.

The robot described sounds more like WALL-E, with a brain and heart full of feelings, right? Well, for John B, “nowadays in the world of sidewalk robotics, there are no fully autonomous robots who can act like WALL-E.” This means that the sidewalk robots out there, are doing just some of their tasks autonomously.

Autonomy at Kiwibot

For John and Carlos, autonomy in robots is a software system's capacity to manage a robot automatically, without humans in the loop, so that it can go from point A to point B safely.

This mobilization must avoid collisions, either with dynamic objects such as people or with static objects, such as walls and chairs. Also, autonomous robots must follow the existing sidewalk rules, more precisely at intersections, respecting certain regulations for crossing. As we said before, all of the above must be done without human intervention.

In our case, the Kiwibots are autonomous in some of the following scenarios. John and Carlos classify the following as the four main problems to solve in terms of autonomy:

  1. The robot stays centered on the sidewalk.
  2. The robot identifies the crossroads.
  3. The robot identifies when it is safe to cross the road.
  4. The robot crosses the road.

Currently, the Kiwibots are able to roll autonomously and stay centered on the sidewalk without human assistance, and that is what we call C2C (Corner to Corner). That means we are covering a part of the big problem and our robots only need human intervention to cross the road, which is the scenario with the most risks.

However, John and Carlos are working hard with their team to make our Kiwibots more autonomous in more and different scenarios.

Work hard, love your job harder

So, how did John and Carlos get into the field of autonomy in the world of robotics? And why did they choose this instead of spaceships? Or instead of searching for alien life? Quick answer: because robots are way cuter than aliens.

\

No alt text provided for this image


The real answer for John is “working in robotics autonomy means trying to solve a problem that no one has solved before, and that feels amazing.” For him, autonomy is a problem that is still state-of-the-art and requires further and constant research, and most importantly, passion; “when you work on something you are passionate about, you grow faster.”  

John feels that he never stops learning at Kiwibot. Every day something new appears, which he seeks to solve not only from his own eyes but from the eyes of all team members. Also, he thinks it is inevitable to feel challenged and love his job at the same time.

On the other hand, Carlos thinks “the most exciting thing about robotic autonomy is that nothing is written, we are building the future.” For him, working at Kiwibot means defining the future of last-mile delivery robots and building the technology that will mark the next years.

So what will the future bring? Let’s see…

The future of autonomy

At Kiwibot, we are working on different projects in terms of autonomy. Here’s a sneak peek: we will have a stronger and more powerful system to make sidewalk navigation more efficient, with an improved version of the C2C. Also, our Kiwibots will be able to cross roads without humans in the loop.

Finally, for John and Carlos, autonomy is not only about the robot itself, the real challenge is to make all the last-mile delivery infrastructure automated. Not only driving the robot, but also the whole process, from when the order is received, to when it is loaded and then delivered to the customer. All of it should be autonomous, that’s the real challenge and that’s what we are working on, to have autonomous and safe operations.

Last but not least, if you have any doubts about working on robotic autonomy, John has something to tell you, “There are so many projects for the future, so many things we are developing that are going to change and revolutionize the world in the next five years.” So, what are you waiting for?


Should John’s and Carlos’ story inspire you or if you are interested in working at Kiwibot, check our careers in robotics. We are hiring!

By
María José Guzmán R.