Safety-Based Control Switching for Collaborative Driving

Kurzfassung

The Managed Automated Driving (MAD) concept introduces a dual-domain architecture in which both the vehicle and the infrastructure provide parallel control paths. Safe operation in this setting requires a switching-based control approach for collaborative driving that reliably detects degradation in the active control path and transfers control authority before the situation becomes critical. Existing approaches lack an explicit, measurable degradation criterion based on observable output signals. The novelty of this work is a structured concept for KPI-driven supervisory switching in MAD systems. A forward-simulation prediction module based on the full nonlinear single-track bicycle model runs in parallel and produces predicted state sequences for both control paths. In the current prototype, four trajectory quality KPIs are derived directly from real-time vehicle states and serve as the decision criterion for a directed, one-way transfer of control authority. Connecting the validated prediction output to the supervisor is identified as the primary next step. A prototype implementation in MATLAB/Simulink validates the concept across three evaluation scenarios. The results confirm that the supervisor reliably detects controller degradation and recovers trajectory tracking quality after the mode transfer.