Situ-Centric Reinforcement Learning for Recommendation of Tasks in Activities of Daily Living In Smart Homes

Thumbnail Image
Date
2018
Authors
Oyeleke, Richard O.
Yu, Chen-Yeou
Major Professor
Advisor
Committee Member
Journal Title
Journal ISSN
Volume Title
Publisher
©2018 IEEE
Authors
Person
Research Projects
Organizational Units
Organizational Unit
Journal Issue
Is Version Of
Versions
Series
Department
Computer Science
Abstract
In this paper, we propose a new recommendation system that considers the changing human mental state, behavior and environmental contexts to provide guidance to persons diagnosed of mild cognitive impairment (MCI) or Alzheimer’s in performing their activities of daily living (ADLs). The recommendation system is based on the Situframework that represents an activity as a sequence of situations, and a situation is a 3-tuple where M denotes human mental state, B is the behavior context and E is the environmental context. More so, it uses an automaton to model an activity thereby generating corresponding task sequences. Interactions between the inhabitants and the sensor networks and the activity being performed are represented as an information space. It then uses a modified model learning algorithm as its decision making tool to recommend appropriate actions or tasks sequence to the inhabitant in a situation of uncertainty caused by episode of confusion or memory loss to ensure he reaches his goal. Consequently, this helps ensure that safety of the person is not compromised when he makes a poor decision regarding the sequence in which ADL tasks are to be performed. We use a single activity class dataset generated from a smart home for 50 independent learning experiments for three (3) ADLs case studies. The accuracy of the sequential-tasks recommendations was found to be high. We further evaluated the risk sensitivity of the recommendation system using the action selection probabilities and corresponding reward points associated with performing a task. The results show that the proposed recommendation system is risk averse.
Comments
This is a manuscript of a proceeding published as R. O. Oyeleke, C. -Y. Yu and C. K. Chang, "Situ-Centric Reinforcement Learning for Recommendation of Tasks in Activities of Daily Living In Smart Homes," 2018 IEEE 42nd Annual Computer Software and Applications Conference (COMPSAC), 2018, pp. 317-322, doi: 10.1109/COMPSAC.2018.10250. Posted with permission.

© 2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.
Description
Keywords
Citation
DOI
Copyright