Game theoretical approach on vehicle-cyclist interaction analysis.

Abstract: Traffic conflict between cyclist and vehicle is typical phenomenon at interaction, and some microscopic models have been proposed to describe the cyclists’ and vehicle drivers’ behaviour. In this thesis, the interactions between vehicles and cyclists are formed as games. The subjects involved in the interaction are modelled to pursue their maximum rewards. This thesis assumes that all game players have an objective of travelling by a steady speed subject to safety constraints. With the assumption of being non-cooperative, static, strategic and complete information game, the players apply strategies satisfying the Nash equilibrium. Four different models are formulated by adjusting the number of model parameters. Based on the modelling framework, a case study is carried out concerning cyclist-vehicle interaction at a Swedish unsignalized roundabout. Using data collected by video cameras, the model parameters are estimated by the maximum likelihood estimation (MLE) approach. Furthermore, one best model is selected according to the performance results. The selected model is examined by data collected afterwards, and the results of validation indicate that this framework is effective in capturing bicycle-vehicle interaction behaviour. Using the better model identified, simulation is carried out to analyse the influential factors on the yielding probabilities for both cyclists and vehicle drivers.