⚑  Palo Alto, California, United States 

Jeff leads the odometry and localization team within Vehicle Perception, a group focused on enabling vehicles to understand their precise position and movement using signals from state-of-the-art sensors. Jeff shares insights into the foundational assisted and automated driving capabilities under development, and also explains how Woven by Toyota’s emphasis on collaboration and craftsmanship guides their work as they strive to deliver driving experiences that give people true peace of mind.

Q: What are you building?

We're building autonomy functionality in such a way that we can expand capability over time. The way that I think of the role of my team is answering the question “Where am I?” for the vehicle, and there are two ways we answer this question. The first is in a relative sense, such as “Where I am now relative to where I was previously?” and “How am I moving?” or “How fast am I travelling?” We call this the odometry system.

The other way we answer “Where am I?” is asking “Where am I on a map?” and “How does this map information augment other structures that I’m inferring about the environment?” — “what’s the speed limit on this road?” and “Where am I along a route that I want to travel?”.

Q: How did you get started in this field? 

I’ve always enjoyed math, but engineering is a way to ground math in real-world problems — things you can understand.

I’ve always loved working with sensor data because sensors magically take real-world information, like the light intensity off-ground, some inertial signals such as acceleration and angular rate, or position with respect to a constellation of satellites in outer space. You get to take all of these signals and fuse them together to create something meaningful to an algorithm.

Q: How does our shared culture with Toyota influence the way you approach this work?

Toyota cares about safety and craftsmanship, and I think that is well aligned with how I feel about rolling out this technology. We want to be careful and we want to be deliberate.

Q: What is it like to engineer within Toyota’s ecosystem?

Toyota is super complex — it’s huge. We don’t own the entire production platform; we own one chunk alongside dozens of other groups, so it’s about understanding how they play together and how to integrate it. It’s definitely a different engineering skill set than sitting down and writing software.

Q: How is our AI-driven approach advancing AD/ADAS systems?

For a vehicle to make a safe decision – to center it in a lane or to change lanes accurately – we need to know where the lanes are. In traditional AV systems, information about lane geometry or lane topology, how lanes are connected, what traffic lights control which lanes, where there is a stop sign, how lanes are connected across intersections — all that information is embedded in an HD map. The reason that this information is embedded in an HD map isn’t because we like building maps, it’s because this is a really hard scene-understanding problem to solve. 

What we can do is build a geometric map, send it to human annotators, and they can mark up where there’s lanes, where there’s traffic lights, how those traffic lights are connected to lanes and how lanes are connected across intersections. 

ML and AI over the last several years have moved so fast that now it is realistic to solve these problems online. And we don’t have to keep maps up-to-date and we don’t have to build maps everywhere. Honestly, I’ve been really excited to see the progress that we’ve made.

Q: What makes a great teammate in your engineering group? 

I love building things. I’m an engineer because I want to build things. I want to take some math, I want to take some data, and I want to take some actuators and make something move. I also love working with people — we build a better system because we’re a better team. Having a desire to want to build a cool system, and having a desire to want to build that as part of a team, are probably the two attributes that I’d really like to encourage folks to bring here.