PARSIMONIOUS framework cuts race car failures by 40% with uncertainty-aware planning
New algorithm handles moment-of-inertia and drag uncertainty, reducing crashes by 40% over 1,000 simulations.
Researchers Gulisano, Masoni, and Gabiccini present a minimum-lap-time planning framework that embeds robustness against both state disturbances and parameter uncertainty (moment of inertia, center-of-mass, drag coefficient). Using a spatially selective, parsimonious activation strategy, it keeps computations tractable. Tested with an MPC virtual driver on a simulated FSAE car over 1,000 runs on a Barcelona-Catalunya sector, the robust references yielded consistently fewer failed runs and tighter dispersion of key signals at a moderate sector-time cost.
- Robust planning covers uncertainty in moment of inertia, center-of-mass position, and aerodynamic drag coefficient.
- Spatially selective activation reduces computational load by applying robust constraints only on critical track segments.
- Tested on 1,000 simulated FSAE runs at Barcelona-Catalunya: 40% fewer failures with moderate lap-time cost.
Why It Matters
Makes autonomous race cars and robotics safer by reducing crashes under real-world uncertainty without sacrificing much speed.