In this research loop detector data patterns significantly associated with crash risk on the freeway are explored. Historical crash and loop detector data from the Interstate-4 (I-4) in Orlando have been used for this research. The possibility of using real-time weather information as part of the crash prediction system on I-4 is investigated. A "rain index" based on the archived rain data from certain locations around the freeway corridor is developed. It is then used, in addition to the loop data, in one of the existing crash prediction models. It was shown that including the "rain index" improves the classification performance of the model marginally. Seemingly Unrelated Negative Binomial (SUNB) crash frequency models were used to identify the geometric design elements of the freeway that could be used as inputs to the real-time crash prediction models. It was found that the presence of ramps is one of the most significant geometric design elements affecting crash risk on the freeway. A section of the I-4 corridor was simulated using the microscopic traffic simulation model PARAMICS. Applying real-time crash prediction model(s) to the simulation data, a measure of crash risk was estimated. Intelligent transportation system (ITS) strategies involving variable speed limits (VSL) were then employed in the simulation environment to examine their impact on the measure of crash risk. It was shown that at moderate-to-high speeds VSL may be used to improve the safety situation on the freeway in real-time. (Author/publisher)
Abstract