Waymo: Pioneering the Future of Self-Driving Cars

Aahaan Maini
9 min readApr 12, 2021

--

You are on your way to the grocery store. You open your phone, enter a location and call a car to come to pick you up and take you there.

Once the car arrives, you climb in the back, buckle up and say hi to the driver. But there is no response. So you ask again: “Hey, do you know where I’d like to go?” Still no response.

You then lean over and see… no one. The driver’s seat is empty. Completely empty. Then just as you start to turn around, the doors close and the car starts moving.

Now you are in full panic mode. You’re hyperventilating and regretting every bad decision you’ve ever made in your life. You are praying that everything is going to be okay and every time you look out the window you get more freaked out.

After a couple of minutes, you control your breathing and work up the courage to remove your hands from your eyes. Only now is when you start to grasp the amount of incredibly technological achievement around you.

You are sitting in a freaking fully self-driving car. None of the assisted driving stuff where you need your hands on the wheel, but full self-driving. Without a driver in the car. You start to grasp how crazy the vehicle around you is.

You are riding in a self-driving car… in 2021. 2021. Today. This is not a concept or idea of the future, no you are actually sitting in a car without a driver, in 2021.

As you reach your destination, you keep pinching yourself to make sure this is not a dream. The car slows to a stop. The doors open.

“Thank you for riding Waymo One. I hope you had a pleasant experience.” You finally get the reply you’ve been waiting for.

The Company Behind the Madness: Waymo

The concept of Waymo began as a self-driving car project in 2009. After joining and graduating X, the moonshot factory, it became Waymo in 2016. It was initially owned by Google under X (also known as Google X) but is now owned by Alphabet Inc.

X is a one-of-a-kind company, focused on moonshot companies. A moonshot company is a company cultivating drastic growth in the tech industry. These companies are focused on 10X improvements to current solutions and discovering massive breakthroughs.

We know that Waymo was one of the most successful moonshot companies to graduate out of X, but what makes it a moonshot?

The Formula for a Moonshot (Credit: X)

Each moonshot has three basic parts: the problem, the solution and the technology that makes the solution possible.

The Problem — Human Drivers

Apart from bringing everyone's vision for the future of transportation to life, Waymo was developed with the goal of solving one of the most important problems we face today: vehicle accidents.

I am sure we all know the devastating impact that car and other vehicle crashes can have, whether you know someone it has affected or you have seen it on the media, the bottom line was that there was something going wrong in the process.

Turns out, the answer was us. Human drivers and them not paying attention. To put this problem in perspective, let's look at some of the statistics.

  1. 35 million lives are lost yearly due to vehicle crashes
  2. 94% of crashes in the United States are due to human error.

Waymo’s mission is to make it safe and easy for people and things to get where they’re going. They are building Waymo Driver to save thousands of lives now lost to traffic crashes.

The solution seems simple, eliminate the human drivers, eliminate human error and save hundreds of thousands of lives. In principle, yes, but developing the technology to do so is where the real moonshot comes in.

The Technology — Waymo Driver

Waymo Autonomous Vehicle Animation (Credit: Waymo)

The Waymo Driver is unlike any other self-driving technology out there because of something called the SAE J3016 Levels of Driving Automation.

Credit: SAE International

Most other companies advertise their vehicles as self-driving when really they are talking about driver assist, which describes levels 0, 1 and 2 on the index. In these scenarios, the human driver must still be in the driver's seat, with both hands on the steering wheel, while still staying alert and being ready to take over control of the car at a moment's notice. For example, Tesla’s Autopilot is only level 2 autonomy.

The Waymo Driver is level 4 autonomy. Level 4 means that the car will not require the passenger to take over control in any scenario. No hands on the wheel, no being alert, nothing.

According to Waymo’s website, the passengers don’t even need to know how to drive.

The Waymo Driver follows a very intuitive 3-step process to carry out its autonomous driving.

Step 1: Data Gathering and Classification

The Waymo Driver uses a combination of custom maps and real-time information gathered through sensors to determine its surroundings.

In terms of the sensor data, the system has an integrated system of various different devices: Lidar, Cameras, Radar and of course, a computer to interpret the information.

LiDAR stands for Light detecting and ranging is a sensing method that uses lasers to measure distances to objects and map the objects around them. How it works is that light is converted to laser pulses, which the car then sends out thousands of pulses per second. The pulses bounce off nearby objects and then return to the car. By calculating the speed of the pulse and the time taken to return, the computer can create a 3D map of the surroundings. Lidar can produce a 360-degree view around the car, of objects a few centimetres all the way to 60 metres away.

The main importance of LiDARs is that they provide a detailed, 3D map of the car’s surroundings. It provides shape and depth to the objects, roads, buildings and more.

Visualization of Lidar Data (Credit: Automotive World)

Cameras are used with object detection algorithms to classify all the various objects the car sees into usable data for the algorithm. The cameras are also placed for 360-degree view and are adapted for identification in various light and climate conditions.

Once the camera has gathered the data, an object detection algorithm, which involves the use of computer vision and nerual networks is used to classify the objects into labels like pedestriations, cars, signs, and more.

To learn more about how computers classify objects, click here for the basics and here to see how he actually builds the model to perform the classification.

Visualization of Object Detection

Waymo uses Radars to detect an object's speed and confirm its distance. Radars are especially important during rain, fog, nighttime and they can ensure the car has an accurate view of everything around it because it is also has a 360-degree view.

Radars work very similarly to LiDARs because LiDAR is fundamentally a variation of a radar. They send out pulses of millimetre-frequency radio waves, which then bounce off objects and can be used to calculate other objects’ distance and speed. This is vital for the next step in the process.

Once we put everything together, the car looks something like this:

How Self-Driving Cars See (Credit: The Economist)

Step 2: Prediction

Now once we have all the data, we need to predict what each object will do and what trajectory it will follow. Driving is an extremely complicated process, with hundreds of different objects, each moving in its own way.

The Waymo Driver uses the information it has gathered in step 1, along with its over 20 million miles driven in the real world (more details below), to predict where other objects (like pedestrians, cars, etc.) would go, how fast they would move, and the way they move.

It predicts all the possible paths each object may take, in less than a second.

Step 3: Path Development and Driving

The car then takes its custom maps, all information it has gathered from the devices, along with all of the possible paths each object may take, to plan the best route to take. Not just where to go, but the exact lane, angle, distance, trajectory to take it closer to its destination.

For the final step, the path the Waymo Driver translates all of the information about its path into usable data like speed and steering maneuvers needed to follow the path, and it carries it out.

The Solution — Waymo One and Waymo Via

Employing its truly groundbreaking Waymo Driver technology, Waymo has two major services that it performs to reduce the number of car accidents per year: Waymo One™ and Waymo Via.

A Waymo One Vehicle (Credit: Waymo)

Waymo One™ is a ride-hailing service that delivers customers from one point to another, in a fully autonomous Waymo vehicle. It uses the Waymo Driver technology to transport people only in the East Valley of Phoenix, Arizona.

Waymo Via is Waymo’s autonomous trucking service, employing, you guessed it, Waymo Driver. It is currently operating 8 heavy-duty trucks in California, Arizona, Texas and New Mexico.

The Waymo City

Waymo is currently only operating in very specific areas and Waymo One™ is only operating in a suburb of Phoenix Arizona.

Waymo has over 600 minivans that are used on a daily basis by the inhabitants of the town to move around from one place to another. The process is extremely simple, you order a car from the Waymo app, it picks you up and takes you to your desired location.

You know the 20 million miles driven I mentioned earlier, this is where it has all been tested, since October 8th, 2020 when Waymo One™ became accessible to all the residents in Phoenix. Read more here.

All of the data collected through these rides has helped the model improve and improve, and it will continue to do so for as long as it lasts.

Now it's time to address the elephant in the room, why is Waymo only operating in one city? Turns out, the answer ties back into its level 4 autonomy on the index. The answer is the same reason why it isn’t at level 5.

The answer has to do with Waymo’s custom maps. As you know, Waymo uses its custom maps as a big part of its program and for the Waymo Driver, and these maps need to be developed. This means driving around, recording every little detail of the area, which has currently only been done in Phoenix.

If you refer back to the index, the example feature of level 5 autonomy, states that the vehicle can drive anywhere in all conditions, whereas right now the Waymo Driver can only drive in locations it has mapped out previously.

Waymo is changing the future of transportation forever. Remember when we were dreaming about self-driving cars 50 years ago? Well thanks to Waymo, they are pioneering turning this dream into a reality.

Thanks for Reading!

Feel free to connect with me on LinkedIn if you enjoyed this article, or email me at aahaanmaini@gmail.com if you have any questions/want to discuss more! If you want to stay up to date on all of the cool things I am doing, subscribe to my monthly newsletter!

--

--

Aahaan Maini

16-year-old ML dev currently building Circulate to tackle the blood shortage in India