Generating Scenarios of Freeway Reliability Analysis

Abstract

The freeway reliability methodology proposed for the Highway Capacity Manual, which is based on SHRP2 L08 methodology, produces an approach to scenario generation that can result in several thousand scenarios to be evaluated to estimate travel time reliability. This large number of scenarios can result in cumbersome user input, a demanding computational burden, and more important, extensive challenges posed when trying to error-check and interpret individual scenarios or to calibrate the model on the basis of real-world observations. This paper presents a novel scenario-generating methodology that accounts for multiple operating conditions. The objective of the proposed approach is to increase the quality of each scenario to make it more representative of the expected congestion patterns on the freeway. This paper shows that the new approach estimates reliability performance measures more accurately than current methods, while reducing the number of scenarios significantly. Thus, the new approach results in a more direct interpretation of results, while simultaneously relaxing many assumptions in the present approach to scenario generation and decreasing biases and errors. The proposed approach uses three core mathematical schemes: (a) a deterministic mathematical model for demand generation and scheduled work zones, (b) a Monte Carlo simulation for incident and weather events, and (c) an optimization algorithm to maximize similarities between the generated set of scenarios and the population of all scenarios. A comparison of results between the proposed method and the SHRP 2 Project L08 approach confirms that the proposed approach yields a higher level of accuracy in matching observed freeway reliability performance measures.