Sidewalk roughness data classification by cluster analysis

Abstract

As sustainability becomes an important aspect of city and transportation planning, individuals are encouraged to choose walking as a mode of travel. In Atlanta, 8.6% of the population under the age 65 are individuals with disabilities, and the sidewalks are indispensable for their mobility; however, many of these sidewalks do not meet the standards of Americans with Disabilities Act (ADA) and need to be repaired. Georgia Institute of Technology researchers developed the Semi-Automated Sidewalk Quality and Safety Assessment System to evaluate and prioritize sidewalk repair projects. This thesis extends the sidewalk roughness levels estimation developed in previous studies. The objectives accomplished in this study are: 1) comparison of the performance of two different tablets for collecting sidewalk vibration data, and 2) exploration of the effects of other related factors on the sidewalk roughness classification result. To accomplish the first goal, k-means cluster analysis is conducted using RMS acceleration data (sidewalk vibration data) collected by Toshiba ThriveTM and Getac® Z710 tablets to classify sidewalk roughness levels. The chi-squared test and Wilcoxon signed-rank test are used to compare the clustering results from the tablets’ RMS acceleration data. This thesis also explores the potential benefits of using other related factors (such as jerk and gyroscope data) on sidewalk roughness classification result. The analytical results show that both tablets generate essentially the same sidewalk roughness classification results and that the sidewalk roughness classification results are dependent of the types of input data used in clustering.