Experimental Study for Crowdsourced Ride Quality Index Estimation Using Smartphones

Abstract

The use of smartphones to collect roughness measurements or ride quality has become popular in recent years, owing to the potential for substantial cost savings in data collection. Hurdles to widespread adoption of these techniques include the quality, uncertainty, and variability of the measurements. The objective of this paper is to evaluate the multifactor effects of collecting ride quality measurements from smartphones and how crowdsourcing using measurements from a population of different vehicles can be used to overcome or mitigate the effects of less resolute and more variable measurements. This investigation was carried out using two experiments. First, a full factorial design of experiment (DOE) was developed to evaluate the multifactor effects on ride quality measurements using the average rectified slope (ARS). Second, a custom DOE with the objective of analyzing the ARS from a population of vehicles from different classifications was carried out. The results from the first experiment suggested that individual factors contribute to statistical differences in ARS measurements. However, the second experiment showed that when looking into a population of vehicles with randomly sampled factors, the ARS measurements converge, and the statistical analysis showed no significance. This approach can be successfully implemented using a crowdsourcing approach where the focus is to analyze a population of vehicles instead of individual measurements.