Onboard measurement and warning module for irregular vehicle behavior

This paper develops an onboard decision module for issuing appropriate warnings when the equipped vehicle's traveling trajectory is irregular. Potential vehicle accidents exist if the driver is distracted by fatigue, drowsiness, food, phone use, and talking or is under the influence of alcohol or drugs, etc. Usual freeway accidents include lateral and rear-end collisions. Detecting vehicle behavior is a good way to measure the security of driving. This paper presents two modules. An Unexpected Lane Departure Avoidance Module is utilized to prevent lateral collision. This module issues warnings when the vehicle approaches an irregular departure from the middle of the lane. Radial basis probability networks are applied to distinguish between normal lane change and lane departure. A Rear-End Collision Avoidance Module is used to issue warnings to avoid longitudinal accidents. This module reflects driver perceptions of environmental influence with a warning value and the warning threshold by neural networks and fuzzy membership functions, respectively. The proposed modules are satisfactory according to simulations and field tests.