Building a driverless race team from scratch

I came back from six months at Audi in Ingolstadt with one thought fixed in my head: there is an astonishing amount of technology packed into a single vehicle, and I wanted to be near it. I enrolled in HorsePower Hannover more or less the week I got back.

The timing was luck. Formula Student Germany had just introduced Driverless — an entirely new competition category, running for the first time that season. There was no playbook for it, no prior driverless team, nobody who had done it before. I joined halfway through the racing season and, because I arrived with robotics experience from earlier projects and from the automated guided vehicle work at Audi, I was made Technical Lead for the driverless car.

The season the CV actually names

My second season is the one that appears on my CV, and the line there — leading a group of twenty-four students — does not describe the actual problem. Everyone from the first driverless season had finished their studies or moved on. What I inherited was not a team. It was a competition entry, a racecar built to be driven by a person, and a room that needed filling.

I want to be precise about that racecar, because “from scratch” is easy to overclaim. It was the previous season’s electric car — their design, their build, made for a human driver — handed to us with a single brief: make it drive itself.

What that brief contained is worth spelling out, because “conversion” sounds lighter than it was. A car built for a person has a person doing the steering and the braking, so both had to be replaced with something a computer could command: we developed and integrated an electronic steering system, and a pneumatic emergency brake capable of stopping a car with nobody in it. Then the senses — a lidar, two cameras, and an inertial navigation system good enough to tie them together — and an Nvidia compute unit quick enough to run the deep learning models on track rather than afterwards. All of that draws power the original car had never budgeted for, so the low-voltage battery had to be redesigned. And once you have added that many components, the wiring loom is no longer the wiring loom: we rebuilt the entire harness.

The chassis and the drivetrain were theirs. Nearly everything between a sensor and a wheel was ours.

We were still never standing on our own: we sat inside HorsePower’s wider organisation and used the electric team’s production facilities and IT infrastructure throughout. The driverless roster had to be rebuilt from scratch. The organisation around it did not, and a good many people who do not appear anywhere in this story are the reason.

What was genuinely ours to solve was the rest of it, and the work stopped being only technical. I recruited the members. I found the sponsors. I planned the budget and ran the finances. I organised the logistics for testing and for competition. And I still owned the technical direction of the car.

That year taught me the thing I have relied on in every role since: on a programme with no precedent, the engineering is rarely the binding constraint. Getting the right people, the money and the parts into the same place at the same time is. It is also where I learned that a team you build yourself has no institutional memory to fall back on — everything your predecessors knew has to be re-derived, or written down well enough that the next generation does not have to.

Four motors, one control problem

By the third season I stepped back from the captaincy. Two years of giving Formula Student everything had left my Mechatronics degree waiting patiently in the background, and it was time. But I could not quite leave, so I stayed on to write the driving dynamics control software: launch control, and torque vectoring across four electric wheel-hub motors.

Four independently driven wheels is a genuinely lovely problem. You are no longer asking how much torque the car should make, but how to distribute it — corner by corner, millisecond by millisecond — to put the power down without spinning up a wheel or pushing the car wide. It was the first time I had a control system where the interesting question was allocation rather than magnitude.

The conversation that became a thesis

That work is also what opened the next door, though I had no idea at the time.

We were testing the launch control at Ehra Camp — Volkswagen’s Formula Student warm-up, an invitation-only event open only to teams VW sponsors — when I ended up in conversation with a Volkswagen engineer about what the controller was doing and why. That conversation turned into my Bachelor thesis in VW’s longitudinal and lateral dynamics department, working on a pre-development hybrid powertrain concept, and eventually into a patent for detecting wheel spin.

I have thought about that since. The opportunity did not come from an application form. It came from being somewhere specific, doing work I could explain properly when someone asked.

Formula Student is where the systems-engineering instinct started — not because anyone taught it, but because a driverless racecar punishes you immediately and publicly for every interface you did not think about.