Abstract
© 2018 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
The concept of activities of daily living (ADL) has for many years successfully been used in a broad range of health and health care applications. Recent hardware and software developments suggest that the future use of ADL will not only benefit from the transition from manually created ADL logs to automatic sensor-based activity recognition and logging but also from the transition from manual inspection of ADL sequences to their automatic software-driven analysis. This ADL sequence analysis software will be core part in mission critical systems, like ambient assisted living, to detect for example changing health status. Therefore, proper testing and evaluation of this software is mandatory before its deployment. However, testing requires data sets that include normal ADL sequences, hazards, and various kinds of long term behavioral changes; which means it might require weeks or even months to monitor individuals to capture such ADL sequences. Thus, collecting such data sets is very costly, if feasible at all; and very few data sets are available on-line. Therefore, we present an approach to create the necessary data sets for testing through simulation. The simulation of ADL sequences is based on existing ADL sequences and uses probabilistic activity instigation and durations with a novel concept called activity rules to create data sets for proper testing. Activity rules are used to model how individuals resolve activity conflicts. We implemented these concepts as a discrete event simulator, called ADLSim. The evaluation of ADLsim shows that the simulated ADL sequences are realistic and able to capture the variability and non-predictable behavior found in the real world, and that activity rules can impact simulation results significantly.