Fuzzy modeling of drivers' actions at intersections

Advanced Driver Assistance systems (ADAS) are systems that assist the driver during the driving task. This technology has great potentials in improving driver and traffic safety. It is very important for an ADAS to predict human drivers' behaviors at urban environment to avoid crashes. Because of the complexity of human-vehicle interaction, it is difficult to obtain an explicit model for analyzing the drivers' behaviors. Instead, models are developed for various driver decisions and driving scenarios (such as lane change decisions and intersection scenarios) which can then be integrated using switch models. Intersections are one of the major scenarios that require special attention in driver behavior modeling. This paper uses Takagi-Sugeno as a data driven technique to model and predict drivers' behaviors at intersections. In the proposed technique, a Takagi-Sugeno model is trained for each maneuver using a Gath-Geva fuzzy clustering algorithm. The proposed models are then evaluated with naturalistic real-world driving data collected in urban traffic, and the estimation results are presented. The results suggest that the proposed technique can correctly estimate the drivers' actions at intersections with high accuracy. This technique uses fewer numbers of maneuver models for training that leads to less computational complexity.