Quantitative characterization of AEB pulses across the modern fleet

Abstract

Objective

Characteristics of specific Automatic Emergency Braking (AEB) pulses can result in increased motion of the occupant, which can lead to the occupant being out-of-position such that when a crash occurs protection may be compromised. Quantifying these variations across the modern fleet is crucial to understand the loading environment to which vehicle occupants are exposed. Therefore, we categorized the AEB pulses based on acceleration pulse features such as deceleration magnitude, jerk, and ramp time.

Methods

A total of 2278 AEB vehicle tests (years 2013–2019) were extracted from the Insurance Institute for Highway Safety (IIHS) database and analyzed. The following pulse characteristics were extracted: Jerk (g/s), Ramp-time (s), and Maximum deceleration (g). A subset of tests in which the tested vehicle did not contact the foam target (n = 1665) was analyzed further, with the following additional variables extracted: Deceleration time (s), Steady-state deceleration (g), and Duration (s). Other non-pulse related features were also considered: Test speed (20 and 40 km/h), Curb weight (Kg), and Vehicle Model Year. Using machine learning methods, the pulses were categorized into clusters. One-way ANOVAs for continuous variables and X² for categorical features were used to assess differences between clusters (p ≤ 0.05).

Results

Using the entirety of the AEB vehicle tests extracted (n = 2278), a total of 3 clusters were selected. The three clusters showed significantly different Jerk, Ramp-time, and Maximum deceleration (p < 0.001). Target contact decreased in AEB tests with more recent vehicle model years (rate of contact 66% in 2014 vs 1.7% in 2019). In one cluster, Jerk and Maximum deceleration increased with vehicle model year. Using the subset of tests in which there was no contact with the foam target (n = 1665), 4 categories of pulses were selected. In both sets of clusters, Ramp-time and Jerk showed moderate inverse correlation (r = –0.7), while all other features showed a low correlation.

Conclusions

These results show that AEB technology improved over the years in obstacle avoidance. The identification of AEB pulse clusters is important in order to describe distinct approaches to achieving AEB and to be able to reproduce representative AEB pulses in the laboratory and understand the influences of those pulses on occupants’ motion.

Keywords:

• Pre-crash
• machine learning
• jerk
• ramp-time
• maximum deceleration
• cluster

Acknowledgments

The authors would like to acknowledge the National Science Foundation (NSF) Center for Child Injury Prevention Studies at The Children’s Hospital of Philadelphia (CHOP) and The Ohio State University (OSU) for sponsoring this study and its Industry Advisory Board (IAB) for their support, valuable input and advice. The views presented are those of the authors and not necessarily the views of CHOP, OSU, the NSF or the IAB members. The authors are also grateful to the IIHS for publishing the AEB test data.