Development of PathDetect: Algorithms that Utilize Smartphone Sensor Data to Detect Adverse Pathway Conditions

This project develops PathDetect, a smartphone app with algorithms that passively characterize adverse pathway conditions (APCs) as a wheelchair user traverses pedestrian pathways. As cities work to improve sidewalks and pedestrian walkways, accurate sidewalk assessments can provide important data to address APCs which can have a significant impact on older adults and individuals who use wheelchairs. The aims of this project are to: (1) Develop an obstacle course that simulates APCs that are compliant and non-compliant with accessibility guidelines published by the US Access Board; (2) develop and evaluate classifiers using a support vector machine and supervised machine learning algorithms to detect APCs from smartphone data collected while using a wheelchair; (3) conduct a “round-table” discussion with wheelchair users to collect important user feedback about the PathDetect tool; and (4) develop and characterize a custom smartphone app that collects sensor data and geo-locates APCs.