Novel NMPC design using finite-gain stability for autonomous vehicles
New control method handles unknown disturbances without Lyapunov complexity—tested on automated driving.
Researchers Carlo Novara, Mattia Boggio, Lorenzo Calogero, and Michele Pagone have introduced a novel approach to Nonlinear Model Predictive Control (NMPC) design based on Finite-Gain Stability (FGS) concepts. Traditional Lyapunov-based methods struggle when plants are affected by unknown but bounded disturbances. The team bypassed this by deriving FGS conditions for the closed-loop system, then systematically choosing NMPC parameters (horizon, cost weights, constraints) that guarantee stability while maintaining good tracking performance—even for time-varying reference signals. This provides a principled, optimization-friendly alternative to ad-hoc tuning.
A simulated case study on an automated vehicle's lateral and longitudinal control confirmed the framework's effectiveness. The approach promises safer, more reliable autonomous systems without requiring exact disturbance models. The paper is available on arXiv (2606.18863) under Systems and Control and Optimization and Control subjects.
- Uses Finite-Gain Stability instead of Lyapunov to handle unknown bounded disturbances in NMPC.
- Provides a systematic methodology for choosing NMPC parameters (horizon, costs) that guarantee closed-loop stability.
- Demonstrated on lateral/longitudinal control of an automated vehicle with time-varying references.
Why It Matters
Enables robust, stable autonomous vehicle control without requiring perfect disturbance models—safer real-world deployment.