System Fault Detection in Human-Augmented Automated Driving

Lateral control of a vehicle in the Automated Highway System (AHS) has been formulated and simulated as part of the hierarchical AHS control structure. In that structure, lateral control resides in the vehicle, and is implemented as a closed-loop control system with the lateral deviation of the vehicle from a reference position as a controlled variable and the steering angle or its rate as a controlling input. The PATH AHS scenario has assumed a road reference and sensing system based on magnetic markers equally spaced along a highway lane. Freedom in the selection of the polarity of each magnet has been utilized to embed binary information along the lane, and it has been utilized to embed preview information (e.g. road curvature and superelevation) useful for the on-board controller to enhance performance both in terms of the lateral error and the occupant's riding comfort. Frequency shaped linear quadratic control (FSLQ), nonlinear sliding control, and rule-based control have been investigated for lateral vehicle control for lane keeping. Simulations of and experiments with this automated system have established that it performs quite well (keeping deviations from the lane centerline under 10 cm). This performance was verified in actual vehicle demonstrations during the 1997 NAHSC demonstration in San Diego. There is, however, still a fundamental question left open: how can the human operator (HO) best be integrated into the automated highway system?