Overview Study of Partially Observable Hidden Markov Models for Ambient Movement Guidance Support.

Abstract

The study of ambient movement guidance encompasses a multidisciplinary approach to facilitating and guiding individuals, particularly older adults, within their living environments. This involves integration of ambient sensors, such as motion detectors, cameras, or IoT devices, to monitor the movements and activities of individuals in real time. By leveraging these sensors, the system can predict and anticipate the expected movements of the person, allowing for proactive ambient guidance and support. In addition to ambient guidance, robots can also play a role in leading individuals by interfacing through audio prompts or visual cues through their daily activities. However, despite advancements in sensor technology and robotic assistance, uncertainties persist in the monitoring and prediction of movements. These uncertainties can arise from various sources, including sensor noise, occlusions, environmental changes, and inherent variability in human behavior. Addressing these uncertainties requires probabilistic modeling techniques based on partially observable hidden Markov models (POHMMs) and various of its extensions such as POMDP, to effectively capture the dynamic nature of movement patterns and incorporate uncertainty into the decision-making process. This paper presents a detailed overview study of probabilistic framework and how its various interpretation can be used in developing an ambient movement guiding system for supporting individuals, particularly older, in support of ageing-in-place paradigms.