How does a Kiwibot navigate autonomously?

November 30, 2023

If it blows your mind when you see a robot rolling along the street, and you wonder how it does it all by itself, this story is for you. The best part is you don't need to be a robotics expert to understand how a Kiwibot navigates autonomously. We’ll lay it all out for you!

First, let’s back up a bit. Remember our last blog about autonomy? We told you Kiwibots could stay centered on the sidewalk without human assistance and we called that C2C (Corner to Corner). Also, we shared a sneak peek of the next step in the advancement of autonomy. Guess what? It is a reality now! Things move quickly here at Kiwibot!

We have been at the forefront of innovation to achieve a high level of autonomy in our robots. After months of development and iteration, today the Kiwibots count on a high-driving automation system capable of being self-sufficient in most of their operations. This new platform makes our robot the safest and most efficient in the company’s history.

A Kiwibot is now able to autonomously reach the pre-established destination while solving problems in various scenarios: rolling along the sidewalk, staying focused on its path, avoiding collision with objects, people, and animals, and identifying street crossings.

Wow! But what does it all mean?

The levels of autonomy

Before we dive into the specifics, have you ever wondered about the concept of autonomy?

Our AI&Robotics engineers define autonomy in a very simple way:

A software system capable of perceiving the world, knowing where the robot is, planning where it should go, how it can reach its destination, and controlling the robot’s movement automatically. This, without humans in the loop, so it can go from point A to point B safely.

However, due to the relative newness of the robot delivery industry, for official purposes, Kiwibot and other companies with the same business model stick to the definition and levels of autonomy used in the automotive industry defined by The Society of Automotive Engineers (SAE).

The SAE defines six driving automation levels to describe the capabilities of a system and the role of the user of the vehicle. Although these definitions are made for road vehicles, we present an extrapolation of these categories for the case of our technology.

  • Level 0 - No automation: The teleoperator has to perform all the driving tasks related to steering and speed control. However, the robot may have sensors that can help it perform some activities. In our case, this could be a distance sensor that alerts our teleoperator of a probable collision.

  • Level 1 - Teleoperator assistance: The robot has the capability to control the steering OR the speed (not both of them simultaneously), while the teleoperator performs the other one. The robots that only have level 1 automation require supervision at all times while performing the steering or speed control. Examples of these driver assistance systems in cars could be forward collision warning (FCW), lane-keeping assistance (LKA), and cruise control systems.

  • Level 2 - Partial driving automation: in this level, the robot is able to control the steering AND the speed at the same time. However, the teleoperator has to supervise the robot at all times due to the absence of environmental and event awareness systems. All our robots have already achieved this level of autonomy, so when you see a robot on a sidewalk, it’s probably driving itself.

  • Level 3 - Conditional driving automation: Same as level 2 but, in some zones, our Kiwibot can also make informed decisions for itself under normal operation conditions. They could be dodging obstacles or planning a new route when it is not possible to take the assigned one. Even so, the supervisor is needed at all times to deal with some edge cases where the Kiwibot itself will request the supervisor’s intervention.

  • Level 4 - High driving automation: This is the highest level our Kiwibots have achieved. A level 4 robot is capable of controlling all the relevant aspects of driving when a set of conditions are met (such as good connectivity, GPS signal, and a map that allows safe navigation), making informed decisions by itself, and dealing with edge cases thanks to previously defined routines. However, the supervisor is needed when the conditions are not fully met.

  • Level 5 - Full driving automation: This level is highly desired but very difficult to reach because it requires a robot to be capable of all the previously described operations in a sustained and unconditional way. All decisions are made by the robot, failures are self-corrected, and a supervisor is no longer required at any time.

Thanks to our latest developments, the Kiwibot’s autonomy went from level 3 to level 4. Level 4 or High driving automation refers to self-driving robots that do not require a driver at all times. Now, the Kiwibots can accomplish their tasks autonomously and only need supervision for very specific high-risk moments.

The technology behind it

If this is starting to sound impossible or even magical, remember there is sophisticated technology involved.

Our robots rely on GPS technology, a high-tech satellite solution that provides high-accuracy locations for self-driving machines. They also employ data from camera sensors combined with AI techniques to avoid obstacles and leverage a navigation system that generates multi-kilometer routes that make it possible to reach a pre-defined goal.

The robot receives a point on a satellite map within the campus limits and navigates to it autonomously.

This works in suitable, pre-mapped locations, using a combination of location systems that are supported by perception-based systems. This means we utilize sensors to pinpoint the robot's site and also use cameras and other sensors to further verify and enhance its location accuracy.

How does it work?

After that thorough explanation, let's move on to the practical application. How does all of this make sense during delivery operations?

Once you order your lunch or latest cravings, the autonomous robots generate multiple paths to reach the pre-established destination and alert if an unexpected obstacle or event occurs on one of them. Remember, there is a team onsite to assist when necessary.

However, there is no need to worry. When we carry out technological developments, we always take efficiency into account, but pedestrian safety is our number one priority on campus and in cities.

“At Kiwibot, cutting-edge technology rolling through the campus will benefit the community by providing efficient delivery services and a source of inspiration for students. Their presence and purpose are to open minds and generate new, sustainable ideas that can improve their own and their community’s life,” said Felipe Chávez, co-founder and CEO of Kiwibot.

The first campus to experience the Kiwibot service with the new high-driving automation robots was Loyola Marymount University. The rest of the fleet is ready to roll out on the campuses and cities we serve.

Now, you have an understanding of autonomous delivery robots and can better appreciate the effort behind the Kiwibots when you order your favorite food or beverage. This collaboration between humans, robots, and software is changing the world, one delivery at a time.

Special thanks to the AI & Robotics team and Alejandro Serna

Maria Jose Guzman Rodriguez