Fourteenth International Conference on The Constructed Environment

Abstract

Older adults’ walking could be easily limited by stressful interactions within the constructed environment. As older adults’ physical capabilities reduce due to aging, walking infrastructures designed for average-person pose great possibilities to cause stressful interactions, which potentially discourage older adults’ walking activities. In this regard, finding better routes with less stressful interactions could promote older adults’ walking by increasing their self-efficacy. The recent integration of AI-based data analytics and wearable sensing has advanced our understanding of the stress during daily trips. This knowledge can be leveraged as valuable input in recommending future routes with more satisfactory experience. Despite this potential, existing studies have primarily focused on identifying individual distress or geolocating stressors. To expand AI data analytics and wearable sensing in route recommendation systems, the authors propose wearable sensor-based framework to assess walking route stress. The study proceeds as follows: (1) apply the individual distress classifier to detect high stress from physical and physiological signals; (2) measure collective distress by aggregating the subject’s high stress data; (3) build a multi-task learning based route-level distress model via collective distress measurements. To test this framework, mean absolute error was measured between the model predicted route stress and self-reported score using 66 subjects’ daily trips data. The results demonstrate that walking route stress could be measured within 10% error range. The proposed framework amplifies the ability of wearable sensors to assess the overall stress level along the route, which could be the basis of building routing systems to improve older adults’ walking.