UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Modeling lateral interactions between motorcycles and vehicles in mixed traffic using four approaches of survival models Essa, Ali


There is growing research interest in evaluating the safety of motorcyclists because of the increasing motorcycling global population coupled with the risks motorcyclists are exposed to as vulnerable road users. An important safety concern for motorcyclists is their lateral interactions with vehicles where a collision avoidance maneuver is needed because of the small lateral separation between vehicles and motorcycles. This study investigates the lateral interaction between motorcycles and vehicles by modeling the critical lateral distance (CLD) between them. The analysis utilized a dataset of motorcycle and vehicle trajectories collected from an urban road network in Athens, Greece. To model the CLD and relate it to various dynamic behavioral and traffic variables (e.g., speed, acceleration, volume, yaw rate), four approaches of survival models were applied and compared. These approaches include fully parametric, fully parametric with Gamma frailty, semi-parametric and machine learning (DeepHit) survival models. The results showed that the DeepHit model outperforms the other three models in terms of the model’s goodness of fit. However, the fully parametric with Gamma frailty model can provide more insights, as it considers the distribution of the CLD and quantifies the influence of behavioral and traffic exposure variables on the probability of lateral interaction. For example, the fully- parametric with Gamma frailty model indicates that the lateral interaction probability increases at higher motorcycle speeds, higher vehicle speeds, higher motorcycle volumes, lower motorcycle yaw rates, and lower relative motorcycle-vehicle decelerations. The results of this study can help quantify and hence mitigate some of the risks that motorcyclists are exposed to, with the overall goal of improving their safety.

Item Media

Item Citations and Data


Attribution-NoDerivatives 4.0 International