Traffic Conflict Identification and Validation using Deep Unsupervised Learning

Surrogate safety measures can provide fast and pro-active safety analysis and give insights on the pre-crash process and crash failure mechanism by studying traffic conflicts. Traffic conflict can be identified by the presence of evasive actions or the amount of temporal (spatial) proximity measures like time-to-collision (TTC). However, it is not enough to use only one kind of measures in some scenarios and it is hard to set a threshold for those measures. Moreover, validating surrogate safety measures by connecting them to crashes is still an open question. We first study the problem of traffic conflict identification. We proposed a method to identify traffic conflict by learning the representation of TTC and driver maneuver profiles with deep unsupervised learning and clustering the representations into traffic conflict and non-conflict clusters. We first trained a transformer encoder to encode sequences of surrogate safety measures into some latent space with unsupervised pre-training. Second, we identified informative clusters in the latent space by calculating the statistic summaries and visualizing trajectory pairs of each cluster. Some clusters are interpreted as traffic conflict clusters because they have small TTC, large deceleration rate and intertwining trajectories and they can be further interpreted as rear-end or angle conflicts. Moreover, the identified traffic conflicts contain critical conditions from the two vehicles in an interaction and one vehicle perceives them as abnormal and takes evasive action to avoid crashes.Secondly, we study the problem of traffic conflict validation. We proposed a method to connect surrogate safety measures to crash probability using probabilistic time series prediction. The method used sequences of speed, acceleration and time-to-collision to estimate the probability density functions of those variables with transformer masked autoregressive flow (transformer-MAF). The autoregressive structure mimicked the causal relationship between condition, action and crash outcome and the probability density functions are used to calculate the conditional action probability, crash probability and conditional crash probability. The predicted sequence is accurate and the estimated probability is reasonable under both traffic conflict context and normal interaction context and the conditional crash probability shows the effectiveness of evasive action to avoid crashes in a counterfactual experiment