For my third year at the EPFL Rocket Team, I started a new research project to replace our passive rocket fins stabilization with thrust vector control (TVC), thus allowing us to control the rocket trajectory and explore propulsive landing technologies.

After a semester of research in Autumn 2020, I recruited a team of 12 students to work on the TVC mechanism and test bench, and help me develop the Guidance, Navigation, and Control (GNC) algorithms I was working on as part of my thesis.
To control the thrust direction of the rocket engine, we developed custom jet vanes, which are small aerodynamic surfaces that redirect the flow of the engine. We simulated, manufactured, and tested many different shapes and materials to have the right thermal and aerodynamic properties. In the end, we chose graphite for its resistance to heat and low density.

A custom PCB was developed by Iacopo Sprenger to integrate a RaspberryPi 4 into our modular flight computer. This gave us a lot more processing power and lets us experiment with advanced GNC algorithms.

To control the vehicle, we used Model Predictive Control (MPC) running on the Raspberry Pi at 50Hz to directly control the actuators. Another MPC was used at a lower frequency to generate optimal reference trajectories for the controller. Finally, we used an extended Kalman filter to estimate external perturbations and model mismatches.
Most of these algorithms were developed by Raphaël Linsen and tested on our simulator, then deployed on our custom drone. Check our recent paper to ICRA for more information on this project!