In recent years, the traffic accidents rate and fatality are decreasing year by year. In comparison, the accident rate and fatality caused by two-wheeled vehicles are not decreasing trends. It is also a problem that the development of the safety systems for the two-wheeled vehicles is insufficient compared to that of the four-wheeled vehicles. Therefore, this study has the purpose modeling to predict the collision avoidance ability in case of a risky situation by using driving behaviors that can be obtained in real-time. For the experiment, a dynamic riding simulator that can control rolling motion was constructed, and the experiment was conducted with 18 test subjects (Mean age = 21.83, S.D. = 1.34). In the experiment, the driving behaviors of each emotional state were investigated based on an emotional model consisting of two axes of valence and arousal with sound stimulation and driving conditions. Driving behaviors were quantified using lateral control ability, head motion as confirmation behavior, and emotional state. The correlation between driving behaviors and collision avoidance ability was investigated. Lane position, one of the indicators of lateral control ability, has a quadratic functional correlation of R2 = 0.568, which is more correlated than other indicators. Moreover, multiple regression analysis was conducted using driving behaviors to predict overall collision avoidance ability. As a result, a model was constructed using driving behaviors with real-time measurement, to predict the rider’s collision avoidance ability when risky situations occur (R2 = 0.685, R2 adj = 0.655).
As deep learning methods in image recognition have achieved excellent performance, researchers have begun to apply CNNs(convolutional neural networks) to automated driving. However, the explainability for the decision making of automated driving is highly desired. In order to trust the model in automated driving, visualization methods have become important for understanding the internal calculation process of CNNs. Therefore, in a previous study, we proposed a method to evaluate the visualization performance of CNN models by using a mathematical model instead of a human driver to generate a dataset that can determine the ground-truth point in images. However, the reliability of the proposed method for validating the visualization performance was not provided. Therefore, in this paper, we verify the proposed method through two experiments to demonstrate the task-dependent performance and visualization performance during training. The reliability of the visualization performance has been demonstrated through experimental results. Therefore, we proposed an evaluation method for visualization performance in automated driving systems.
This paper describes a mathematical model that considers the effects of thermal stress and strain caused by thermal expansion and contraction during cooling and the effect of bending by welding electrodes. As a result of predicting the effect of the initial gap between the steel plate and the welding electrode, the so-called clearance, using the model, when the clearance is large, the strain generated at the time of springback is reduced, and it is shown that liquid metal embrittlement cracking is unlikely to occur. This prediction corresponded well with the experimental results. Various effects other than clearance were predicted by the model.
Being able to generate adequate test scenarios in order to validate autonomous driving functions is of paramount importance for the deployment of autonomous vehicles. However, asserting the completeness of generated test cases and coverage of all possibilities has proven to be very difficult. Previously, few attempts that tried to generate representative possibilities have all been based on the designer ′ s experience and thus inherently incomplete. This paper proposes a formal system inspired approach that solves this issue. With this approach, 2,544 basic use-cases were generated from a completeness perspective. Thereafter, for the sake of accelerating the development work of our autonomous vehicle, we used this asset to clarify which use-cases should be solved and in what order.