An NMPC-Based Control and Decision System as the Enabler for a New HCI in Autonomous Vehicles: Validation in Complex Simulated Scenarios

Automated driving (AD) promises major benefits in road safety, traffic efficiency, and environmental impact, yet full autonomy remains challenging due to highly dynamic, uncertain traffic environments and the need to interact safely with unpredictable human drivers. This work proposes a unified Nonlinear Model Predictive Control (NMPC) framework that functions both as a low-level controller and a high-level decision-making system, enabling automated vehicles to plan smooth and safe maneuvers over a predictive horizon. Unlike many existing NMPC implementations that are constrained to narrowly defined tasks and require scenario-specific parameter tuning, the proposed approach integrates nonlinear vehicle dynamics, multi-vehicle interactions, and safety constraints within a single formulation designed to generalize across complex driving situations. The framework is validated through simulations of challenging real-world maneuvers, including merging in extra-urban and urban settings (with compliance to stop-and-go and traffic regulations) and navigating multi-vehicle roundabouts with dense interactions. Results indicate collision-free operation, smooth trajectories without abrupt maneuvers, and robust adaptability to varying traffic densities and speed profiles.