Abstract Manager

Traditional road safety relies primarily on reactive methods that aggregate crash records over several years. This results in a moral dilemma where a statistically significant number of traffic injuries and fatalities must occur before action can be taken. Conflict-based analysis has recently been used in practice as a proactive approach to safety. However, conflicts remain a surrogate measure that do not necessarily accurately represent the actual crash risk. New statistical advances have been proposed for proactive safety that result in the ability to accurately determine the safety level of a location in real time based on road users’ trajectory data. These new statistical techniques allow for the estimation of real-time crash risk over very short time intervals. This session presents two metrics, the Risk of Crash (ROC) and Return Level (RL), were proposed that allow for the real-time estimation of crash risk. As more data is becoming available in real-time, and processing power is dramatically increasing, the profession must quickly adapt to leverage real-time data availability and apply it to real-time safety analysis. Connected vehicle data, autonomous vehicle data, computer vision, and cell phone data have become increasingly available and provide great potential for road safety applications. Combining the new technique with these new data sources, several applications have been developed that would allow for real-time adaptive decision-making and real-time safety optimization. For road agencies, this technique may be used to dynamically identify hazardous locations based on real-time data. By treating safety as a dynamic measure that changes temporally, new solutions may be proposed to substantially reduce collisions. Additionally, these new metrics may be used for the real-time optimization of safety using traffic signal timing, turn movement changes, dynamic speed advisories and limits, and intelligent dynamic fleet routing. These topics and their potential for implementation will be discussed during the session. Thus, this approach marks a departure away from the reliance on long-term collision data towards real-time approaches with a wide variety of applications in the near future.