Analysis and comparison of two task models in a partially observable Markov decision process based assistive system

Abstract

People suffering from dementia experience difficulties during their daily self-care activities. The resulting loss of independence makes them rely on caregivers to go through their daily routine. However, such reliance on caregivers may conflict with their need for privacy. Hence, there is a need for technology that can provide assistance automatically. In the field of artificial intelligent assistive technology, the module in charge of automatically guiding users during a task is called Task Manager. This paper compares two task modeling approaches in an assistive system named COACH (Cognitive Orthosis for Assisting with aCtivities in the Home), which was designed to provide guidance to older adults with dementia during the handwashing task. The results obtained show how implementing a suitable Task Modeling approach led to 180.4% improvement of appropriately timed prompts provided by the system.