Existing methods for bottleneck detection can be categorized into two: methods based on stochastic analysis and methods based on data-driven analysis. The stochastic methods are accurate in estimating bottlenecks in long term, ignoring the current improvement opportunities, while the data-driven methods tend to do the opposite. In this paper, we develop an optimal policy to integrate the two methods based on Markov decision theory. The characterization of the optimal policy is provided. In addition, to implement the policy, the optimal frequency for carrying out bottleneck analysis is investigated. Numerical experiment is performed to validate the effectiveness of the optimal policy and compare it to the existing methods.

This content is only available via PDF.
You do not currently have access to this content.