With advances in vehicle modeling and control, intelligent vehicles can increasingly utilize the full vehicle capabilities should it be necessary for safety. This paper presents a shared control approach capable of operating in the open-loop unstable state space of a vehicle. First the Maximal Phase Recoverable Parallelogram is defined as the state space where a control action exists that can stabilize the vehicle. An online approach to constructing a subset of the Maximal Phase Recoverable Parallelogram that avoids spinning out past a desired sideslip is presented. This then forms a safety envelope for a Nonlinear Model Predictive Control framework for shared control. The shared control formulation is designed to match driver inputs, unless inputs lead to safety violations such as a spin or track bound violation. If intervention is needed, the controller seeks to follow the drivers intent in a safe manner. Results for a full scale experimental vehicle executing circular steady-state drifting demonstrate the ability of the controller to operate in the open-loop unstable drifting regime and limit intervention unless there is a risk of spinning out or violating track bounds. This shared control approach is a step towards assisting a driver in using a vehicle’s full capabilities even in extreme maneuvers. READ MORE