Driver Intervention Detection via Real-Time Transfer Function Estimation

Currently, the driver plays a crucial role in the safety of autonomous vehicles, functioning as the fall back for vehicle control systems. Fast and accurate detection of driver intervention is therefore of importance. In this article a novel driver intervention detection method for automated vehicles is presented and tested. Non-critical transitions are considered, excluding safety related applications. The transfer function between the electric power steering torque and steering column angle is estimated by perturbing the steering system with a known disturbance. This estimated value is used to detect whether a driver is intervening. The detection algorithm has been tested in simulations using a four degree-of-freedom vehicle model. The parameters of this vehicle model have been obtained via frequency response measurements performed on a test vehicle. Secondly, the performance of the algorithm has been tested with on-road measurements. The results show that driver intervention can be successfully detected within 0.4 seconds. The performance in terms of true and false detections has been analyzed, and the presented solution is shown to be robust to measurement noise and road disturbances.