Evaluating Shared Control in Real-World Conditions

This paper presents an openly documented real-vehicle evaluation of an adaptive shared control system that integrates a nonlinear Model Predictive Controller (NMPC), a Driver Monitoring System (DMS) and a haptic Human Machine Interface (HMI). While similar concepts have been explored by several industrial players, detailed performance data and transparent methodologies remain scarce in the public domain. This work aims to help fill that gap by providing insights into the behavior and integration of such systems under real driving conditions. The system establishes an active safety envelope for manual driving scenarios, employing a shared control strategy that activates during potentially hazardous situations caused by driver misbehavior or external road challenges. A key contribution of this work is the real-world validation of the approach in a vehicle with human drivers, demonstrating how shared control can be effectively deployed beyond simulation. The proposed framework integrates three elements: 1) an NMPC-based shared controller that shifts control authority across three different modes (Electronic Power Steering, Shared Control, and Fully Automated) by adjusting its weights; 2) a DMS that triggers mode switching; and 3) a haptic feedback interface that communicates system state, transitions, and driver intervention requests. The controller enforces a lane-keeping safety envelope and intervenes only when necessary. Results from real-vehicle experiments across two use cases show that the system enhances safety by mitigating lane-departure risks and improving driver attentiveness.