Resampling Methods for the RPGD Optimizer
Bachelor's thesis (5.75/6.00) — sampling-based nonlinear MPC for autonomous driving and racing.
My Bachelor’s thesis investigated novel resampling strategies for the Resampling Parallel Gradient Descent (RPGD) optimizer — a sampling-based, non-linear model predictive control scheme used for autonomous driving and racing.
The study explored how different resampling methods affect the convergence and robustness of the MPC, evaluated in simulation and on the F1TENTH platform.
- Co-supervisor: Prof. Dr. Emilio Frazzoli — IDSC, ETH Zürich
- Co-supervisor: Prof. Dr. Tobias Delbrück — Institute of Neuroinformatics (INI), ETH Zürich & UZH
- Grade: 5.75 / 6.00
- Keywords: autonomous vehicles, model predictive control, LiDAR, simulation, F1TENTH