Alphabet Subsidiary Waymo Shares New Details About its Self-Driving Class-8 Trucks

author: Eric Walz   

Alphabet subsidiary Waymo, which spun out of Google's self-driving car project, has been actively developing autonomous vehicles for the past decade. In addition to its self-driving cars, Waymo has been working on self-driving trucks since 2017 with plans to launch a commercial freight delivery service called Waymo Via.

Today in a blog post, Waymo shared new details about its advanced driverless trucks that are poised to transform the shipping industry, making freight delivery much safer and more efficient. So far, the autonomous Class 8 trucks have been tested on public roads in a wide variety of cities and environments in California, Georgia, Arizona, New Mexico and Texas. 

For starters, Waymo's driverless trucks use the same core technology suite as Waymo's self-driving passenger vehicles, which the company calls the "Waymo Driver." The Waymo Driver is the foundation for all of Waymo's autonomous vehicle development. 

In the blog post on Tuesday, Waymo wrote that by using the same technology as its passenger vehicles, each vehicle platform can benefit from the 20 million self-driven miles on public roads and over 15 billion miles in computer simulation that the company has completed to continuously improve its software. 

With this extensive onroad experience, Waymo has now unlocked freeway driving capabilities for our entire fleet. Building a truck that can drive itself on the highways is a challenging endeavor for Waymo and every other company working on the technology. 

Some of the challenges facing human truck drivers are piloting an 80,000 pound fully-loaded semi-truck at 65 mph, merging into fast-moving traffic, dealing with blind spots, maintaining lane position, constantly checking the mirrors to keep an eye on the trailer as well as other vehicles. So driving a big truck is no easy task.

As a result, more than one in three long-haul truck drivers have experienced a serious crash during their careers, according to data from the U.S. Bureau of Labor Statistics.

However, Waymo's self-driving technology can bring a level of safety to the industry that cannot be achieved by human drivers. But Waymo is not simply replacing truck drivers with computers. Instead the company is tapping into the wealth of knowledge gathered from professional truck drivers and incorporating it into its software stack.

As part of its testing program, Waymo said it partners with test drivers who are trucking industry veterans with more than 20 years of invaluable expertise to help improve the Waymo Driver technology stack.

"By working with the engineering teams and sharing all about truck behavior and the rules of the road, I'm helping the Waymo Driver see and learn what I have. It's my job to impart the lessons I've learned the hard way, so that the Waymo Driver is the safest it can be," said Jon Rainwater, an experienced truck driver who provides instruction to the test drivers for Waymo Via.

"That is the largest impact I can have—knowing that society will benefit from my lived experience for years and years to come."

Each of Waymo's trucks are outfitted with the same hardware that most self-driving vehicles rely on, including camera, lidar and radar systems. Rather than installing the hardware on a production vehicle that is already manufactured, Waymo is working directly with truck manufacturers and OEM partners to integrate its Waymo Driver seamlessly into their vehicles during production. 

This includes optimizing the self-driving system and adapting it to vehicles its installed on, such as setting up the cameras to eliminate potential blind spots for each specific model. Waymo has also been developing its own proprietary hardware, including lidar sensors.

However, one of the more important features of Waymo's self-driving trucks is having a long-range perception system that exceeds that of passenger vehicles. 

Compared to cars, Class-8 trucks are designed to operate on freeways and their mass and weight means that it takes longer to stop, up to the length of two football fields. The trucks are also slower to accelerate than passenger cars.

To accommodate for these differences, Waymo increased the number of onboard sensors. The most notable modification is utilizing two perception systems as opposed to the single perception dome that's found on the roof of Waymo's self-driving cars. The dual perception domes help increase rear visibility by reducing blind spots created when hauling a 53-foot long trailer.

The sensor suite on Waymo's self-driving trucks offer 360 degree coverage for increased safety.

Using Machine Learning to Improve the 'Waymo Driver'

For self-driving vehicles, machine learning is an important tool that can be used to predict the behavior of other road users. To safely navigate, Waymo's autonomous vehicles rely on highly complex, high-definition maps and a constant stream of vehicle sensor data. However, this data alone is not enough to make predictions about what might happen on the road, according to Waymo.

In its latest blog post, Waymo stressed the importance of machine learning in addressing freeway driving requirements for perception, prediction, and path planning. Waymo said that its fine tuning the algorithms its uses for its self-driving cars to better support operating the large and heavy trucks at highway speeds. 

To address these issues and make better predictions Waymo developed a new machine learning model it calls "VectorNet", that provides more accurate behavior predictions while using less compute power.

Another important task was the development of a longer range perception system for the trucks. The Waymo Driver for trucks can identify objects ahead at greater distances, which allows it to respond earlier and maneuver more smoothly if needed.

Some of the other challenges unique to long-haul trucks are navigating traffic metering lights when merging onto a freeway and moving over a lane when another vehicle stops on the shoulder to give them room. In addition, the driverless trucks must also contend with frequent highway construction zones and median crossovers that temporarily direct traffic to oncoming lanes. 

Training the Waymo Driver to navigate a variety of these scenarios is challenging, but necessary. The company said it leads to additional freeway driving capabilities for the entire fleet.

Waymo developed a machine learning model called "VectorNet" that provides more accurate behavior predictions.

Ensuring the Reliability of a Big Truck's Self Driving Systems

Another challenge for Waymo or any developer of autonomous vehicles is ensuring exceptionally high reliability from every part of the trucks' self-driving system. 

The Waymo Driver goes through rigorous testing and validation processes to ensure the highest level of safety and reliability. This process includes base vehicle and hardware level reliability and durability testing, as well as validating software through simulation, then finally testing the software on public roads using trained drivers.

The simulated environment can recreate any scene a truck driver might encounter in the real world, including the behavior of other drivers on the road that frequently don't follow traffic laws.

As Waymo continues to develop the hardware and software components of the Waymo Driver to power trucks, the technical progress achieved will carry over into all other vehicle platforms, including its passenger cars.

Waymo is not the only company working on self-driving trucks. Others include California startup TuSimple and Swedish company Einride.

Self-driving truck technology may soon make transporting goods less expensive, safer and much more efficient. With over a decade of experience in the development of autonomous vehicles, Waymo continues to lead this effort. 

Despite the economic fallout due to the coronavirus crisis, venture capital continues to flow to companies like Waymo that have the potential to change the future mobility landscape. In 2020 alone, Waymo has raised $3 billion in new investments.

Eric Walz

Originally hailing from New Jersey, Eric is a automotive & technology reporter covering the high-tech industry here in Silicon Valley. He has over 15 years of automotive experience and a bachelors degree in computer science. These skills, combined with technical writing and news reporting, allows him to fully understand and identify new and innovative technologies in the auto industry and beyond. He has worked at Uber on self-driving cars and as a technical writer, helping people to understand and work with technology.