Kaosisochukwu Egbuonu, "Bayesian Network-based Intention Estimation of Traffic Participants ", Technical University of Darmstadt, 2024.

Autonomous vehicles are seen as a great source of hope when it comes to improving general road traffic safety. In order to achieve this and prevent potential accidents, they are required to predict the motion of surrounding traffic participants. Generally speaking, the movement of traffic participants is driven by hidden intentions such as taking a left turn or going straight at an intersection. Therefore, this work aims to develop an approach to estimate the route intentions of traffic participants in road traffic scenarios. For this purpose, the data provided by the Waymo Open Motion Dataset was utilized to design a dynamic Bayesian network as model for urban traffic scenarios. Additionally, two novel white-box models, the route transition model and the maneuver model, were developed and embedded. The intention estimation was conducted by means of the particle filtering algorithm. Eventually, two versions of the developed intention estimation algorithm, one with and the other without the route transition model, were evaluated and compared. The results show that the intention estimation algorithm without the route transition model is capable of correctly estimating the intention of traffic participants. However, vehicle interactions remain a central challenge. Beyond that, the presented approach serves as a starting point for further efforts dedicated to the intention estimation task.