‘The next frontier in robotics’
Zoox principal software engineer Olivier Toupet on company’s autonomous robotaxi technology
Olivier Toupet has dedicated his career to making the impossible possible — on Earth and in space.
“I've always worked on futuristic technologies,” said the principal software engineer for Zoox, an independent Amazon subsidiary that is building a self-driving, purpose-built, electric robotaxi fleet. “It’s what keeps me motivated.”
His career is certainly proof of that.
Before joining Zoox in November 2021, Toupet spent nine years working at NASA’s Jet Propulsion Laboratory (JPL), where he supervised the Robotic Aerial Mobility group. The group “develops novel robotic technologies for unmanned aerial vehicles (UAVs) and provides robotic solutions tailored to unique challenges associated with aerial robots—with a focus on guidance, navigation and control to expand autonomy and mobility capabilities.”
For Toupet, a job that involved aerial robots was a perfect fit for a kid who grew up fascinated by flying.
Toupet was born and raised in the South of France, where his passion for aeronautics was fueled by his family’s frequent travels. “My mother worked for a French airline, so we got to fly a lot,” Toupet recalled. “And I remember my dad teaching me the names of all the airplanes at the airport and wanting to be a pilot when I grew up.”
He studied at the prestigious French Institute of Aeronautics and Aerospace in Toulouse before moving to the United States to pursue a Master’s in Aeronautics and Astronautics at the Massachusetts Institute of Technology (MIT). There, he was able to pursue his dream of working on cutting-edge aircraft. Boeing Phantom Works funded his graduate research into autonomous path-planning and decision-making for unmanned helicopters.
After graduating in 2006, he spent a 10-month stint working on some of the first aerobatic helicopter drones. He then moved to senior engineering roles with Aurora Flight Sciences (which Boeing acquired in 2007) and Zee.Aero (now Wisk), where he developed artificial intelligence (AI) for unconventional electric vertical take-off and landing (eVTOL) aircraft. Or, as he calls them, flying cars.
“eVTOLs are closer to what I am now doing at Zoox,” explained Toupet, who has a private pilot’s license. “However while flying does come with its own set of challenges, including FAA regulations, the aerial environment is less cluttered and more predictable than driving in a busy city. The contrast is even starker with the Martian environment, where nothing moves except for the rover, at 0.1 miles per hour, making path-planning much easier to perform.”
From flying cars to driving rovers
Toupet joined JPL in 2013 where he developed novel technologies for the Mars rovers, including traction control for Curiosity and self-driving for Perseverance. He received two NASA honor awards for that work, the Exceptional Engineering Achievement Medal in 2017 and the Exceptional Technology Achievement Medal in 2022.
During his tenure at JPL, Toupet drove three Mars rovers — Opportunity, Curiosity, and Perseverance — and flew the Ingenuity helicopter. “It was a truly unique and amazing experience,” he recalled. “There is a lot of specialized training involved — it takes over a year to get certified for both mobility and robotic arm operations. I loved learning how the spacecraft works and how to control it with flight software.”
As part of the Mars rover operations team, he also learned to handle the pressure that comes with being responsible for the safety of a multi-billion dollar national space asset.
“We often had to make critical decisions quickly to ensure we had trustworthy commands to send to a rover during the limited communications window when it’s in line of sight with the orbiter,” he added.
He divided his time at JPL between technology development and robotics operations. That approach enabled him to see the impact of his work in action and better understand the operational context and gaps that required additional research.
“Exploring regions of Mars never seen by human eyes before was extremely cool, of course,” he added. “But what I remember enjoying most was the incredible people from different backgrounds I had the privilege to work with, and the great camaraderie we shared.”
Hungry for his next challenge, in 2021 he pursued what he saw as a once-in-a-lifetime role at Zoox.
“No one has been able to produce a vehicle capable of driving itself fully autonomously in the rich and varied environment of a busy city,” Toupet said. “To me this is the new frontier in robotics not only because it’s so challenging at every level of the robotics software stack — perception, planning, prediction, control — but also because I believe that technology is attainable in the next few years. The self-driving car industry is an incredibly innovative and booming sector, with great potential for making a really big impact on our society in the near future. That’s why I find it the most exciting place to be for a robotics engineer.”
Robotics for a better world
Today, as principal software engineer, Toupet is developing cutting-edge AI algorithms that enable the self-driving Zoox vehicle to understand and make decisions based on its surroundings, and to optimize trajectories to reach its destination safely and comfortably.
Zoox vehicles employ advanced perception systems that, among other things, detect and identify cars, pedestrians, bicyclists, and other mobile elements that might affect the vehicle’s path. The challenge lies not just in spotting those elements, but also in predicting their future motion.
“In a way, it’s similar to what I was just doing at JPL, where I developed the path-planner for the Perseverance Mars rover, but the problem is a lot more challenging for a fast-moving self-driving car in a dense, rich and dynamic urban environment,” Toupet said. “We are using machine learning — specifically, self-supervised and reinforcement learning — to better predict the intent of surrounding vehicles based on past data collected, and to improve the potential actions our vehicle can take.”
Toupet’s team is actively testing their algorithms, experiencing their behavior first-hand by riding in the Zoox vehicles. They’re also using virtual simulations to stress-test the technology, running many instances in parallel on the cloud to drive many more miles than would be otherwise achievable and test for edge cases that rarely occur in the real world.
“Zoox brings together the best of all worlds,” Toupet concluded. “At Zoox, I work collaboratively with some of the most talented robotics experts in the industry and together, we design, implement, and test state-of-the-art AI algorithms running on cutting-edge hardware, that will have a big, near-term impact on society by making public transportation safer, greener, and more enjoyable for everyone.”