Crash recognition algorithm of automatic crash notification system with adaptive discrimination threshold

Abstract

Efficiency and notification accuracy are two critical criteria for evaluating the performance of automatic crash notification (ACN) systems. The discrimination threshold (DT) is used to assess whether a collision accident occurs. Typically, the DT value is determined based on the maximal acceleration peak from multiple road tests. Because an overlarge DT value is unnecessary in most driving scenes and simultaneously adversely affects notification accuracy, a crash recognition (CR) algorithm with adaptive DT is proposed. First, a road–vehicle simulation model is constructed using the CarSim software. Next, the vehicle acceleration data at different driving speeds are obtained based on this road–vehicle model. Subsequently, a correlation model comprising a discrimination threshold value, an international roughness index, and the vehicle speed is developed. Finally, a CR algorithm is designed, in which discrimination threshold values that match the road roughness and vehicle speed are specified. Road and collision tests show that the proposed algorithm can identify collisions and calculate the speed change value more accurately compared with the conventional CR algorithm which has a fixed discrimination threshold.