چارچوب شبکه های بیزی بر اساس پیش بینی سقوط زمان واقعی در بخش آزاد راه اساسی بزرگراه های شهری

The concept of measuring the crash risk for a very short time window in near future is gaining more practicality due to the recent advancements in the fields of information systems and traffic sensor technology. Although some real-time crash prediction models have already been proposed, they are still primitive in nature and require substantial improvements to be implemented in real-life. This manuscript investigates the major shortcomings of the existing models and offers solutions to overcome them with an improved framework and modeling method. It employs random multinomial logit model to identify the most important predictors as well as the most suitable detector locations to acquire data to build such a model. Afterwards, it applies Bayesian belief net (BBN) to build the real-time crash prediction model. The model has been constructed using high resolution detector data collected from Shibuya 3 and Shinjuku 4 expressways under the jurisdiction of Tokyo Metropolitan Expressway Company Limited, Japan. It has been specifically built for the basic freeway segments and it predicts the chance of formation of a hazardous traffic condition within the next 4–9 min for a particular 250 meter long road section. The performance evaluation results reflect that at an average threshold value the model is able to successful classify 66% of the future crashes with a false alarm rate less than 20%.

The manuscript investigates the major shortcomings of the current initiatives regarding predicting crash risk on urban expressways in real-time and offers solutions to overcome them with an improved framework and modeling method. The queries regarding the detector positions from which data will be extracted to perform prediction has been answered by conducting the study on a heavily instrumented urban expressway such as the Tokyo Metropolitan Expressway. The issue of large variable space with small sample size has been addressed by introducing random multinomial logit model to identify and rank the most important variables. The study also underlined that the variables under consideration in real-time crash prediction models are highly correlated by nature and thus the modeling methods employed must be robust enough to accommodate those variables. Moreover, it also highlights that many times surrogate variables may be needed to be used to model such a problem due to lack of data availability. These models may as well need to be updated with the partially available new data on some variables. For this, it is required to be able to update itself in modular way. The model will need to accommodate new variables in future without requiring to completely rebuilding itself. Considering all these specific requirements, this study has introduced Bayesian belief net (BBN) as a modeling method. The proposed model also binds its result with a space (250 meter section) and time (for the next 4–9 min) which will be necessary for researcher involved with countermeasure designing. To elaborate more, when evidences are entered, the model predicts the chance of a traffic condition of a 250 meter long section under consideration in the basic freeway segment to become hazardous within the next 4–9 min. The findings regarding the most appropriate detector position suggest that a detector placed approximately 250 meters downstream from the centroid of the section under consideration can capture the abnormality in the traffic conditions with the highest precision. The second best detector position is a location 250 meters upstream from the centroid of the section under surveillance. The study identifies that the traffic conditions in the upstream and the downstream as well as the difference in the traffic flow parameters in these locations have high impact in precise detection of hazardous conditions. A new variable called ‘congestion index’ has been introduced through this study, too, which was calculated by comparing the instantaneous speed of the stream with the free flow speed at that location. The final variable set used in the model comprised of: congestion index in the downstream and upstream and the difference in speed and occupancy between the upstream and the downstream. The study has also demonstrated how the complexity of the model can be reduced by combining the classical ‘parent divorcing’ technique with conventional BBN. From the performance evaluation point of view, the model has been built following a conservative approach to capture mainly highly hazardous road traffic conditions maintaining a low false alarm rate. This approach is important as the road authorities are expected to take counter measures such as warning the drivers with variable message signs (VMS), controlling the driving speed through variable speed limit (VSL), maintaining the level of congestion through ramp metering and sometimes even with drastic measures such as main line metering and lane change prohibition. Therefore, as the knowledge regarding proactive evasive measures to countervail traffic conditions impending to crash is still at its early development stage, it is expected that actions will be taken only when the risk of a crash is substantially high. Regarding the performance evaluation of the model, it was also taken into account that many a time a crash does not occur even under a hazardous traffic condition due to the skills of drivers and several crashes also take place due to factors that cannot be captured with traffic flow variables. Therefore, the 66% success rate in capturing hazardous traffic conditions with a less than 20% false alarm rate can be considered satisfactory. The study presents a choice for the road authorities through Fig. 5. Decision makers may choose different threshold values based on their needs to decide upon evasive actions. Moreover, the proposed model is robust from the maintenance point of view, too. It is built with data from two detectors and thus it can still yield results even if one detector fails. It is also important to highlight that the model need not be implemented throughout the basic freeway segments. The expressway authorities are not required to layout detectors throughout the length of the expressway either. As it binds crash with both time and space, the authorities can identify hazardous locations on the road with conventional crash prediction models and layout detectors as recommended in this study and monitor those specific areas only. This provides the expressway authorities with higher level of flexibility from the financial point of view. Although the study presents a new framework coupled with modern modeling methods to bridge the gap between conceptual research and practical implementability, it has its own limitations. The manuscript addresses the issue of real-time hazardous traffic condition monitoring for the basic freeway segments only and recommends a new study to develop separate models for the ramp vicinities following a similar framework. The research mainly focuses on timely detection of hazardous traffic condition formation and does not provide much insight on the underlying crash mechanism. The study also does not cover issues related to appropriate intervention design. Nevertheless, the new methodological framework demonstrated in this study can be used as a core and further studies regarding crash mechanism understanding, improving prediction performance and designing counter measures can be conducted around it.