MDP-Based Reliability Analysis of an Ambient Assisted Living System

The proliferation of ageing population creates heavy burdens to all industrialised societies. Smart systems equipped with ambient intelligence technologies, also known as Ambient Assisted Living (AAL) Systems are in great needs to improve the elders' independent living and alleviate the pressure on caregivers/family members. In practice, these systems are expected to meet a certain reliability requirement in order to guarantee the usefulness. However, this is challenging due to the facts that AAL systems come with complex behaviours, dynamic environments and unreliable communications. In this work, we report our experience on analysing reliability of a smart healthcare system named AMUPADH for elderly people with dementia, which is deployed in a Singapore-based nursing home. Using Markov Decision Process (MDP) as the reliability model, we perform reliability analysis in three aspects. Firstly, we judge the AAL system design by calculating the overall system reliability based on the reliability value of each component. Secondly, to achieve the required system reliability, we perform the reliability distribution to calculate the reliability requirement for each component. Lastly, sensitivity analysis is applied to find which component affects the system reliability most significantly. Our evaluation shows that the overall reliability of reminders to be sent correctly in AMUPADH system is below 40%, and improving the reliability of Wi-Fi network would be more effective to improve the overall reliability than other components.