مدل شبیه سازی BP و تجزیه و تحلیل حساسیت تصمیم گیری برای عبور و گردش به راست وسایل نقلیه در تقاطع چراغدار

Inter-crossing behavior model of motor vehicles and bicycles is the key part of micro-simulation for mixed traffic at signalized intersection. The microscopic behaviors of the motor vehicles passing through the bicycle flow at a two phased signalized intersection were analyzed to reproduce the passing behavior of motor vehicles. A BP neural network model was proposed to describe the motor vehicles' passing decision. Based on the field data at two typical intersections in Beijing, the model was validated and compared with the Logistic model. The results indicated that the BP model was more effective than the Logistic model and had better prediction accuracy. First derivative sensitivity matrix of the BP model was established. The sensitivity analysis showed that the most important factor impacting on the motor vehicles' passing decision-making behavior is the gap allowing motor vehicles to pass through. The passing decision-making behavior is the most sensitive to the gap when it lies between 2.76 s and 2.96 s.

Mixed traffic is a typical characteristic of urban traffic in China, especially at road-intersections. Motor vehicles are frequently disturbed by bicycles from different directions at the same phase. These interferences not only significantly decrease the intersection capacities but also increase the probabilities of traffic crash. Targeting on these problems, this study focuses on the crossing behavior between motor vehicles and bicycles and explores the interference mechanism exploration. It also proposes a foundation for the development of a mixed traffic signalized intersection simulation system. The automobile-bicycle mutual crossing through behavior analysis has drawn much attention from domestic and international scholars, and some achievements have been obtained. To summarize, there are two research concepts at present: (1) The probability choice model based on gap acceptance theory is used: Ferrara[1] established a probability choice model of bicycle crossing through bi-directional and four lanes and investigated the stop accepted gap and non-stop

In this study, based on the analysis of the vehicle’s crossing behavior at intersections, the right-turning vehicle’s crossing decision-making model representing the vehicle’s crossing through the straight-moving bicycles on adjacent lanes behavior was established. The proposed BPN model was validated and the sensitivity analysis was given based on the field data collected at two typical intersections in Beijing. The following conclusions are obtained. (1) An alternative model for simulating the vehicle’s crossing decision-making behavior is established. The experimental results show that the mean right recognition ratio of the vehicle’s crossing decision-making model based on gap acceptance and lag acceptance are 98.43% and 97.56%respectively. The proposed model has higher forecast accuracy compared with the Logistic regression model. (2) The impact of four input variables on the vehicle’s crossing decision-making behavior is analyzed quantitatively by calculating the sensitivity matrix of the best BP vehicle’s crossing decision-making model based on gap acceptance. The results show that the greater the speed of motor vehicle before crossing is, the greater the vehicle crossing probability is, so is the gap. The more the bicycles in the conflict zone are, the lower the vehicle crossing probability is, so is the speed of the following-bicycle. (3) The determinant of the vehicle’s crossing decision mode is the gap. By contrast, the speed change of the motor vehicle before crossing has less impact on the crossing decision mode of right-turning vehicles. When the gap lies between 2.62 s and 3.12 s, the gap could have effect on the crossing decision mode. The crossing decision-making behavior is the most sensitive to the gap when it lies between 2.76 s and 2.96 s. BP simulation model of motor vehicles’ crossing decision proposed in this study provides a simulation analysis method for reproduction of the automobile-bicycle interferences at signalized intersection in mixed traffic situation. However, the proposed model is validated and compared with other model based on the field data collected at two intersections only. The universality and stability of the BPN model would be investigated in future study