Skip to content
Maximilian Vogt
← All projects
In progressTODO: e.g. Oct 2025 – present2 min read

GFR Driverless: autonomous stack for Formula Student

The driverless software of Global Formula Racing's Formula Student car: lidar perception, planning and control on ROS Noetic in C++ and Python, developed against simulation before it runs on the car.

Role
TODO: your role, e.g. control and simulation, Autonomous Driving team
Tags
  • Autonomous Driving
  • Robotics

Problem

In the driverless discipline of Formula Student, the car drives a track marked only by traffic cones, with no driver and no map on the first lap. It has to detect the cones, estimate its own position, plan a path and follow it at the limit of grip, with every stage running in real time on the car’s computer.

TODO: Which part of this problem does the team (and do you) work on, and why is it hard? For example, perception at speed, latency between lidar and control, or a controller that stays stable near the grip limit.

Approach

TODO: architecture of the driverless stack and where your part sits.

TODO: The stack is built on ROS Noetic in C++ and Python. Describe the components you worked on:

  • Perception: TODO (lidar cone detection: method, sensor, frame rate)
  • Control: TODO (controller type, e.g. pure pursuit, Stanley or MPC; how it was tuned)
  • Simulation: TODO (simulator, how it is used for testing before track days)

Results

TODO: e.g. cone detection recall at 20 m
—
TODO: e.g. lateral tracking error (RMS)
—
TODO: e.g. stack latency, sensor to actuation
—
TODO: say where each number comes from (simulation or track, which event, which date).

TODO: Results in one or two short paragraphs. What worked, what did not, and what was measured on the real car.

My role

TODO: This is a team project. Name the team size, your sub-team, and what you personally built or owned. Credit the parts other people did.

What I’d do differently

TODO: What you would change in the architecture or the process, and what the next season’s goal is.