Skip to main content

Research Repository

Advanced Search

Investigating the transition from normal driving to safety critical scenarios

Papazikou, Evita

Authors

Evita Papazikou



Abstract

Investigation of the correlation between factors associated with crash development has enabled the implementation of methods aiming to avert and control crash causation at various points within the crash sequence (Evans, 2006). Partitioning the crash sequence is important because intricated crash causation sequences can be deconstructed and effective prevention strategies can be suggested (Wu & Thor, 2015). Towards this purpose, Tingvall et al. (2009) documented the so-called integrated safety chain which described the change of crash risk on the basis of a developing sequence of events that led to a collision. This thesis examines the crash sequence development and thus, the transition from normal driving to safety critical scenarios.
The current research utilises Naturalistic Driving Studies (NDS) and more specifically Strategic Highway Research Program 2 NDS (SHRP2 NDS) data to investigate the crash sequence. Trip-based time series data covering 2.5 minutes prior to the events (crashes and near-crashes) and the corresponding driver and event data were extracted from the SHRP 2 NDS dataset by Virginia Tech Transportation Institute (VTTI). After the data cleaning, matching and transformation process, 773 events with 553 drivers were available for analysis. With the data sampled at 10 Hz, over 1 million data points were included to the final dataset. The analysis conducted in three stages regarding the time sequence in crash development. Firstly, the time period during normal driving stage was investigated, followed by the whole crash sequence and finally, the last time period towards safety critical scenarios was examined.
Safety indicators during normal driving were characterised and functional relationships, providing dynamic thresholds in relation to speed, for departure from normal driving were derived. Longitudinal and lateral acceleration, yaw rate and TTC presented different distributions across gender and age groups. Moreover, relevant safety indicators generated with an empirical process, were employed to examine the whole crash sequence development and recognise deviations from normal driving. The descriptive analysis revealed that yaw rate, longitudinal and lateral accelerations may be feasible determinant of crash risk in earlier stages. Therefore, in the last 30 seconds prior to events, the driver braking, and steering behaviour was explored by extracting events of relevant interest. Examining the events mean values and their duration, thresholds for emerging situations were proposed.
Lastly, TTC values were further investigated and their evolution during crash sequence was analysed by using multilevel mixed effects modelling. According to the random slope model that was estimated, TTC values are affected by vehicle type, longitudinal acceleration, speed, and time within the crash sequence expressed by the timestamp variable.
The outputs of this thesis can be adopted by insurance companies to formulate normal driving profiles for different driver groups, and also, by the automation industry to evaluate or design new collision avoidance or warning systems.

Thesis Type Thesis
Deposit Date Jun 7, 2024
Public URL https://uwe-repository.worktribe.com/output/12037358
Award Date Dec 16, 2019