NETWORK SCREENING IN A CONNECTED VEHICLE ENVIRONMENT

摘要

Transportation agencies are responsible for analyzing crash data to identify hot spots - locations that experience abnormally large numbers of crashes, pointing to potential geometric and or control problems. Current network screening practice involves using information from police reports to determine hot spot locations. There are numerous issues with current practice. First, police reports are often inaccurate with regards to exact location and the cause of the incident. Second, from a statistical perspective, this method requires a large number of crashes to occur before the problem can be recognized, which often takes years. A connected vehicle environment offers the potential to improve this process. In a connected vehicle environment, transportation agencies will have access to more vehicular probe data than ever before. As a result, it may be possible to detect a near miss. Near misses are events during which evasive maneuvers occur and a crash is narrowly avoided. These are not reported to the police so current network screening practice will not have information regarding near misses. Current hot spot location identifier techniques will be applied to near miss event locations Using the CVI-UTC Virginia connected vehicle testbed, data will be collected and near misses will be identified using threshold values determined from a combination of a literature review, analysis of existing data, and contact with professionals in industry. Thresholds will be determined for any data element that may indicate a near miss occurred. This includes longitudinal acceleration, lateral acceleration, yaw rate, and speed in addition to a few variables that indicate the driver's intentions or condition of the vehicle such as use of turn signal or size of vehicle. Following data collection at the UTC Virginia connected vehicle testbed, locations that had near misses frequently occur, will be compared to hot spots indicated by police reports for verification of the proposed method.