MPC New Sampling Method Reduces Costs of Highway Asset Inspections
Posted: Oct 1, 2019
MPC researchers at the University of Utah developed a sampling method utilizing machine-learning techniques to suggest the location and frequency of sampling of roadway assets to determine maintenance needs. This high-dimensional clustering-based stratified sampling method is designed to choose the proper highway segments where the conditions of the sampled assets represent the maintenance performance of the full inventory within the network. By using this method, road management agencies can reduce the resources required for asset inspection. The method can be applied to any high-dimensional sampling process -- for example, in selecting corridor segments, intersections, or traffic assets where multiple types of features such as traffic, geometric design, or assets need to be considered.
Using the inspection records from the state of Utah, the researchers verified that the high-dimensional clustering-based stratified sampling method outperforms the simple random sampling method used by many state departments of transportation.
MPC researchers at the University of Utah developed a sampling method utilizing machine-learning techniques to suggest the location and frequency of sampling of roadway assets to determine maintenance needs. This high-dimensional clustering-based stratified sampling method is designed to choose the proper highway segments where the conditions of the sampled assets represent the maintenance performance of the full inventory within the network. By using this method, road management agencies can reduce the resources required for asset inspection. The method can be applied to any high-dimensional sampling process -- for example, in selecting corridor segments, intersections, or traffic assets where multiple types of features such as traffic, geometric design, or assets need to be considered.