Assistance Method for Merging Based on a Probability Regression Model

Merging behavior requires multiple tasks such as cognition, decision-making, and driving operation. Previously, driving assistance systems, which instruct drivers on making accelerations, have been studied to support the decision-making task. The importance of improving driver comfort with adjusting system variables has been revealed through these studies. The present study aims to propose assistance methods for merging, which decreases driver's workload and difficulty in decision-making. The proposed methods recognize drivers' decision ambiguity using a decision-making model for respective drivers and instruct them on acceleration to decrease the ambiguity. First, we develop a decision-making model to predict where drivers merge based on a logistic function. Furthermore, we propose acoustic assistance methods, which instruct the acceleration and deceleration. The systems continuously calculate the optimal instruction based on driving history from the beginning of the assistance. Driving simulator experiments demonstrated that drivers' workload and decision ambiguity decreased with our proposed methods.