Skip to content
← Build log
Autonomy2026-06-10· 2026 Season

Standing Up the Driverless Program

Year one as Autonomous Systems Lead: reading the driverless ruleset like a spec, architecting a C++/ROS 2 monorepo for the Jetson Orin, and securing the sensors that will let the car see. Target: driverless by 2028.

#ROS 2#C++#Jetson Orin#architecture
AV · FIG.LEAD

Our car does not drive itself yet. That’s the honest headline, and it’s also the point: the team’s goal is a driverless car by 2028, and this season my job as Autonomous Systems Lead (DSO) has been to make sure the foundation we lay now doesn’t have to be torn up later.

Rules first, architecture second

Before writing a line of stack code, I studied the FSAE driverless ruleset and how FSG runs its driverless events, because compliance requirements decide architecture: where the emergency brake authority lives, what the autonomous system master switch has to do, and what the car must prove before it’s allowed to move without a driver. Designing around those constraints from day one is much cheaper than retrofitting them.

The monorepo

The software home for all of it is a C++/ROS 2 monorepo targeting the NVIDIA Jetson Orin. I architected the repository and build infrastructure: package layout, build tooling, and the project structure that lets 10+ software members work in parallel without stepping on each other.

REPO · PACKAGE LAYOUT
RULESET · COMPLIANCE MAP
SENSORS · MOUNT STUDY
DASH · TEAM DOCS

Hardware we don’t have to buy

Driverless perception is expensive, so I ran targeted sponsorship outreach and secured about $30k in sponsored hardware: LiDARs, cameras, and auxiliary sensors. That’s the difference between designing the perception stack around the sensors we want versus the sensors we could afford.

Keeping ten people rowing together

The least glamorous work matters most at this stage: I’m building a team documentation dashboard with visual aids covering sensor mounting, wire harness routing, and subsystem interfaces, so knowledge survives graduation cycles and new members ramp up in weeks instead of months.

Next milestones: perception prototyping on the Orin with the sponsored sensor suite, and simulation infrastructure so the stack can fail safely long before the car does. The build log will follow it, honestly, all the way to 2028.