Riding Performance Quantification Method For Motorcycles In Terms Of Collision Probability Using The Logit Model

In recent years, many Advanced Driver Assistance Systems (ADAS) have been proposed and introduced under the development of sensing technology and the issue of driving safety. But many kinds of ADASs have a specific threshold to control the alarm or some support. This is decided based on the experimental or mathematical calculations in terms of the optimization of the human-machine interface of each system. But almost all of the systems (especially warning systems) have just a single threshold value to issue the warning, and the driving performance of drivers fluctuating in real time is not considered. In this study, we proposed a quantification method of riding performance and performed the logistic regression analysis for the collision prediction model based on riding performance to optimize the warning threshold of ADAS. For this study, 64 test subjects (Mean age = 22.14, S.D. = 3.71) participated in the experiments using simulator. Experiments were conducted for three risk events (left-angle collision when a rider was driving on priority road or driving on non-priority road, and right turning collision) and dummy events with the same road environment without risky situations. We proposed a quantification method of riding performance through the total sum of a product of the generalized value of riding behaviours. We also proposed the logit model, which can be constructed in terms of the collision probabilities and riding performance, which is quantified using our proposed method. In the logit model, collision occurrence was used as the dependent variable and riding performance was used as the independent variable for logistic regression analysis to clarify the condition where the probability of collision increases. Finally, we proposed a concept of the setting method of threshold value for the warning timing of ADAS according to the rider's performance level based on collision probabilities during each riding performance.