Dangerous driving event prediction on expressways using fuzzy attributed map matching

This paper presents a system for predicting dangerous driving events while driving on an expressway. There are three major tasks involved in the prediction system: (1) how to perceive driving events on the input sequence of driving conditions, (2) how to represent driving events, and (3) how to interpret driving events to decide whether or not they are hazardous. A directed acyclic graph, called the attributed driving relational map (ADRM), is introduced to represent driving events. The ADRM chronicles a driving event in terms of driving conditions. The prediction system evaluates the driving event to determine whether it is perilous or not by matching its ADRM against those of known dangerous driving events preserved in a database using a fuzzy attributed map matching technique. The database can automatically augment by including new dangerous driving events that approved any of the predefined danger criteria. A series of experiments with synthetic examples generated by a driving simulator have been conducted to demonstrate the feasibility and rationality of the proposed system.