[Ph.D. Dissertation Defense]

Ph.D.Candidate Eunjeong Ko

Abstract
While traveling on the road, vehicles often encounter sections where different traffic streams converge together due to the geometric structures. These sections require decisions to be made in a short period of time to ensure smooth driving, so vehicle interactions are inevitable, which can lead to safety issues. Therefore, it is important to identify in advance the sections where mandatory lane-changing occur and to clearly understand the decisionmaking of the vehicles. Based on this, this study aims to examine two representative sections where mandatory lane-changing occurs, namely merging sections and roundabouts, and proposes a game-theoretic model to determine vehicle’s strategy by considering probabilistic payoffs. To this end, this study defined the payoff for decision as minimizing travel time or passing time, and formalized the game by a payoff function and matrix. Next, a feasible numerical range was defined to calculate the probabilities of various driving states on the road, and each equilibrium was evaluated for each situation. Finally, an empirical analysis was performed using large-scale aerial image-based trajectory data, and the predicted results were compared with the actual decisions. As a result of the model verification, the predicted results showed a fairly high level of accuracy when compared to actual decisions. This study provides a micro-level exploration of lane-changing behavior, allowing for a realistic understanding of decision-making.

1. Date: Friday, May 31, 2024, 12:00

2. Venue: KAIST Munji Campus Lecture Hall L407