A Detection and Warning System for Unintended Acceleration

This paper presents a data-driven method to detect vehicle problems related to unintended acceleration (UA). A diagnostic system is formulated by analyzing several specific vehicle events such as acceleration peaks and generating corresponding mathematical models. The diagnostic algorithm was implemented in the Simulink/dSpace environment for validation. Major factors that affect vehicles’ acceleration (e.g., changes of road grades and gear shifting) were included in the simulation. UA errors were added randomly when human drivers drove virtual cars. The simulation results show that the algorithm succeeds in detecting abnormal acceleration.