Project Overview
AMZ Racing is ETH Zurich's Formula Student team, and it gave me one of my most formative environments for building autonomous systems under performance pressure. I first worked as an Autopilot Software Engineer on path planning and high-level controls before later leading the controls module.
The project forced every control and planning decision to stand up to tight integration timelines, vehicle dynamics, and repeated iteration across seasons. It was a strong lesson in how to build systems that are both fast and dependable.
Technical Systems
Path Planning
Worked on the planning layer that shaped how the vehicle reasoned about the track and motion objectives
MPC Control
Developed a Model Predictive Controller for the Skidpad event to operate the car near its lateral limits
Module Leadership
Led a four-person controls team toward a reliable and performance-oriented software stack
Cross-Team Integration
Worked across software and vehicle integration while also supporting mechanical and event responsibilities
Key Contributions
Formula Student Driverless is demanding because performance, reliability, and iteration speed all matter at once:
- Path planning and high-level control: Contributed to the driverless autonomy stack during two seasons
- MPC for Skidpad: Built a controller specifically aimed at handling the car close to its lateral limits
- Season review and improvement planning: Studied prior work, identified shortcomings, and proposed concrete upgrades
- Team ownership: Led the controls module and helped maintain a reliable, performance-oriented development direction
Technical Challenges
Operating near vehicle limits: Control work in racing is unforgiving. Small modeling errors and integration issues show up quickly when the vehicle is being pushed for performance.
Iteration across seasons: Competition systems are inherited, revised, and rebuilt continuously. A big part of the work was understanding previous choices, identifying weak spots, and turning that review into better software decisions.
Collaboration under constraints: The driverless stack only works when controls, planning, perception, and vehicle integration all align. That made communication and module ownership just as important as the controller design itself.
Why It Mattered
AMZ Racing gave me a strong foundation in high-performance robotics engineering: how to reason about control under pressure, how to review complex legacy work quickly, and how to lead a small team toward better technical decisions.
It also sharpened a mindset I still value: strong robotics software should be technically rigorous, but it also has to survive deadlines, integration complexity, and the realities of a multidisciplinary team.