For Street Sweeping, the Future is Bright. And Most Likely Driver-Less.

The NiteHawk and Hermes Robotics teams with the autonomous Raptor II sweeper. From left: Tracy Day, Mattia Pelissou, Guilhem Herail, Dale Glubrecht, and Dan Stone.

In 2015, auto manufacturers and tech companies started making big promises about autonomous personal vehicles. Toyota, Honda, GM, Tesla, and Google’s Waymo all announced delivery of self-driving cars by 2020. In 2018, The Atlantic predicted that up to one million people a day would ride in driver-less cars in the next two years. But if you’re one of the 128 million Americans who drove their own car to work today, you know this goal was more than a little ambitious.

There is no question these companies have made tremendous advancements towards autonomous personal vehicles, but many engineering challenges stand in the way. While some are discouraged by the slow progress, innovators still see almost limitless potential in this technological space. And their goals reach way beyond a self-driving taxi.

Driving Forward: Hermes Robotics

Guilhem Herail of San Francisco is one of those innovators. His company Hermes Robotics is working toward autonomous technology, not for personal use, but municipal and commercial applications. Herail and his team believe delivery trucks, waste management vehicles, and even tractors will be driverless long before your Tesla takes you through the Starbucks drive-thru.

“Taxi robots and autonomous personal vehicles still need to prove they are safe,” explains Herail. “These personal vehicles will work eventually in most driving situations. But for driving in a more controlled environment, the technology is already mature enough.”

Municipal and commercial vehicles perform recurring tasks, thus offering the controlled environment in which autonomous technology can safely operate. So which industry in the municipal space will be the first to go driver-less? Herail’s answer might surprise you.

“I think street sweeping could be the first truly mainstream use of self-driving technology.”
Herail and his team at Hermes Robotics are developing a system to enable large, slow-moving vehicles, like sweepers, to operate without a driver. Their goal is to structure the system as an installable kit that can effectively convert vehicles from traditional to autonomous operating systems.

How It Works

Hermes’ technology uses a sophisticated sensing system called lidar, which allows the vehicle to detect and react to objects in its environment. Short for light detection and ranging, a lidar system sends photons from pulsing diodes in all directions around the vehicle. As the photons bounce off objects and back to the vehicle, sensors calculate their distance, using this and other data points to generate a 3D map of the environment. Software translates data from the 3D map and sensors to motors installed on the steering column and pedals, which then direct the vehicle to avoid obstacles without human assistance.

Herail credits robotics engineers and other automotive innovators for making their system possible. “There is a lot of incredible work happening right now in that space. We’re definitely standing on the shoulders of giants,” says Herail.

Lidar works in tandem with GPS systems to create an accurate understanding of the vehicle’s position. “Lidar sensing is an important aspect of our system. It helps us navigate with excellent accuracy, and know for certain where people and things are,” Herail explains.

Sometimes, the system cannot navigate with lidar only, so the navigation is enhanced with GPS. “For example, a vehicle may be in a repeat location, but there are new piles of snow on all sides of the vehicle. With just lidar, the robot sweeper may not recognize its environment. Working with GPS, however, the system can learn that it’s in a repeat environment with new obstacles, allowing it to understand its location.”

This may seem like a lot of technology for machines like street sweepers, but Herail is convinced they are the optimal application.

“The main reason is the speed and time of operation,” explains Herail. “All autonomous vehicles have what is known as an Operational Design Domain, or ODD. These are the defined conditions wherein the vehicle can safely operate. And for sweepers, the ODD is relatively constrained and simple.”

The ODD of a sweeper is much less complicated than the ODD of a personal vehicle. Sweeper trucks run mostly at night, and they visit the same routes over and over. Most importantly, they encounter very few pedestrians. Personal vehicles, on the other hand, navigate very complex and constantly changing environments.

“Think about all the things that can possibly happen while you drive your car,” says Herail. “There are countless problems for the computer to solve in that environment—unlimited possible routes, infinite potential obstacles. But the sweeper environment is a limited problem with a lot of repetition. It’s a problem we can absolutely solve.”

Teaming Up against Trash

Herail was inspired to bring autonomous vehicles into the sweeping world by the thing that motivates just about all sweeping: trash.

“I was walking around my neighborhood in San Francisco, and I couldn’t believe the garbage in the streets. I had seen the sweepers go by, but there was still trash everywhere. We engineers are trained to find the simplest solution, and I thought, how do we decrease the amount of waste on these streets? We have to sweep more often. Automation seemed like the best way to accomplish this.”

Herail has assembled an ambitious team of young engineering professionals to help him achieve this goal.
“I started on my own, retrofitting the first prototype, but the control system (the software’s ability to control the vehicle speed and direction) was way too basic. I knew I needed help if I was going to get this done.”
Herail first enlisted the expertise of Mattia Pelissou, a fellow alum of his Master’s program, as well as a UC Berkeley graduate, to cofound the company and support the engineering side of the business. Together they began honing the technology and recruiting more support. Not long after, they brought Eleonore Jacquemet to the team.
“Eleonore has a Master’s of Science in Robotics from Stanford. She’s really helped us take our technology to the next level.”
Herail named his company Hermes Robotics after the Greek god Hermes. In ancient mythology, it was believed that Hermes and his winged sandals could travel between worlds, defying boundaries and delivering messages swiftly between gods and mortals.

“We believe our company will help take autonomous technology beyond the boundaries of what we now believe possible.”

Growing the Team

The Hermes team partnered with NiteHawk Sweepers of Seattle, Washington, to create their first prototype truck. NiteHawk provided a Raptor II, a medium-duty parking lot sweeper built on an Isuzu NPR chassis. NiteHawk and Hermes also entered an agreement to distribute the startup retrofitting technology in the future.

“The NiteHawk was an ideal test case for this technology. It has a great turning radius, making it very maneuverable in tight spaces. And the quality of the sweep is excellent. We don’t have to worry about blow-out or missing debris and can focus on optimizing the autonomous technology. We know the sweeper will do its job, so we can do ours.”

Currently the Hermes Robotics team is acquiring test hours, analyzing data, and optimizing the autonomous performance of their NiteHawk. They have been able to accrue thousands of test hours through pilot sweeping agreements, servicing many large properties in the Bay Area.

“Right now, we’re sweeping those parking lots every night. We can cover them at about 80% autonomous capacity. Our goal is to get that to 99%.”

The Hermes team continues to work on challenges specific to sweeping, such as training the system to adapt to weather changes.

“Wind is a tricky problem,” says Herail. “A sweeper driver knows when they need to double back to get debris that blew in after the first pass. Teaching a driver-less sweeper to do the same is a challenge, but we’re getting there.”

Rain can also interfere with laser sensing. Herail explains, “Because the system relies on light transmission, distortions happen when the lasers travel through water. Luckily engineers specialized in sensors are working to develop systems that will enable sweepers to generate more accurate sensing, even in inclement weather.”
Several other exciting features are still in development. “In the future we believe we will be able to automate the entire sweeping process. Not just sweeping the route, but tasks like locating dumpsters, emptying the hopper, and even inspecting the lot and delivering reports to property managers.”

Herail has been thrilled by the response from the sweeping industry.

“Some people see sweeping as an industry that isn’t very forward thinking, but everyone we’ve talked to is excited to innovate. People are ready to move ahead with this technology, improve their businesses, and provide better service.”

Herail and his team plan to continue improving the technology and advancing the industry forward. “We have an opportunity to spotlight the sweeping industry as a major driver of innovation.”

He adds with a smile, “and together we can beat taxi robots!”


Piper, Kelsey. 28 February 2020. It’s 2020. Where are our self-driving cars? Available at
Madrigal, Alexis C. 28 March, 2018. The most important self-driving car announcement yet. The Atlantic. Available at
Loesche, Dyfed. 5 October 2017. How Americans commute to work. Available at
National Ocean Service. 26 Feb 2021. What is lidar? National Oceanic and Atmospheric Administration. Available at

No comments yet... Be the first to leave a reply!