A Personalised Formal Verification Framework for Monitoring Activities of Daily Living of Older Adults Living Independently in Their Homes

Ricardo Contreras, Filip Smola, Nuša Farič, Jiawei Zheng · 2025

Read A Personalised Formal Verification Framework for Monitoring Activities of Daily Living of Older Adults Living Independently in Their Homes by Ricardo Contreras, Filip Smola, Nuša Farič, Jiawei Zheng (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-08701.txt.

Introduction

Checking that an older person living alone is safe means noticing when daily routines break. Ricardo Contreras and co-authors build a checking framework with formal proof tools in this 2025 paper. Sensors in the home feed models of each person's routines, built also from interviews, home layouts, and field observations.

Each person's safety needs are written as formal properties in a logic of time. A model checker then proves each property or hands back a trace showing exactly how it failed. The framework was applied to several participants to show it travels across homes and habits. Personal models, the authors argue, beat one-size monitors because routines differ so much. The goal is longer, safer independent living.

Catalog record

CREATORRicardo Contreras, Filip Smola, Nuša Farič, Jiawei Zheng
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE39 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-08701

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

A Personalised Formal Verification Framework for Monitoring Activities of Daily Living of Older Adults Living Independently in Their Homes Thanks: Ricardo Contreras, Filip Smola, Nuša Farič, Jane Hillston, and Jacques D. Fleuriot, are with the School of Informatics, The University of Edinburgh, 10 Crichton Street, Edinburgh, EH8 9AB, UK (e-mail: [email protected]; [email protected]; [email protected]; [email protected]; [email protected]). Thanks: Jiawei Zheng is with DigitLab, University of Exeter (e-mail: [email protected]) and worked on the project while at the University of Edinburgh. Ricardo Contreras, Filip Smola, Nuša Farič, Jiawei Zheng Abstract There is an imperative need to provide quality-of-life to a growing population of older adults living independently. Personalised solutions that focus on the person and take into consideration their preferences and context are key. In this work, we introduce a framework for representing and reasoning about the Activities of Daily Living of older adults living independently at home. The framework integrates data from sensors and contextual information that aggregates semi-structured interviews, home layouts and sociological observations from the participants. We use these data to create formal models, personalised for each participant according to their preferences and context. We formulate requirements that are specific to each individual as properties encoded in Linear Temporal Logic and use a model checker to verify whether each property is satisfied by the model. When a property is violated, a counter example is generated giving the cause of the violation. We demonstrate the framework’s generalisability by applying it to different participants, highlighting its potential to enhance the safety and well-being of older adults ageing in place. 1 Introduction Advancements in medical science have significantly increased life expectancy in recent decades, resulting in a growing population of older adults (OAs) [ 1 ] . As an increasing number of such adults live independently [ 2 ] , greater attention is being paid to ensuring their safety, addressing their care needs and maintaining their quality-of-life (QoL) [ 3 ] . Innovative technologies are increasingly needed to support independent living and enhance the QoL of OAs. Activities of Daily Living (ADLs) are essential self-care physical tasks that an individual performs on a regular basis. ADLs are fundamental for maintaining health and well-being [ 4 ] . Unexpected behaviour during these activities may indicate a decline in cognitive or physical abilities, potentially leading to unsafe conditions [ 5 ] . Moreover, monitoring such activities could be crucial for detecting changes that may impact an individual’s QoL. Sensors provide a powerful means of monitoring ADLs and, by extension, detecting deviations, thus potentially facilitating timely interventions that can safeguard or even improve the QoL of OAs. In this paper, we present a novel framework that combines data from sensors, interviews, home layout and sociological observations (i.e. the systematic analysis of social behaviours and interactions) for representing and reasoning about ADLs in the homes of OAs living independently. The framework leverages unobtrusive, commercially-available sensors to capture ADL events, e.g. opening a refrigerator, closing a shower cubicle door. These events are integrated through an automation platform and stored in a time-series database. To enrich the sensor data, we systematically collect contextual information about daily routines through interviews and sociological observations. This contextual data also enables the customisation of sensor setups and the creation of accurate, person-centred models that account for home layout, privacy preferences, and individual requirements. This work is specifically aimed at OAs living alone in their homes, with the overall goal of identifying unanticipated behaviours over time. By modelling the various actions involved in carrying out ADLs as a finite state model [ 6 ] , we can form an abstraction of the behaviour of the OA. This enables the application of formal methods to verify whether the observed behaviour, captured by a set of sensors, satisfies a set of behavioural properties specific to that OA. These properties are derived from existing general guidelines and from specific requirements of the OA. We use model checking [ 6 ] , a formal verification technique, to verify whether properties hold for the observed behaviour. More specifically, we use the model checker NuSMV [ 7 ] to specify and verify the model, encoding properties in a dialect of Linear Temporal Logic (LTL) [ 8 ] . This study builds upon our previous work [ 9 ] which involved using public datasets [ 10 ] . It significantly extends it through the use of our own data, collected from multiple participants within an OA population in Edinburgh, Scotland. Using the combined sensor and contextual data, we automatically generate models based on the placement of sensors in each participant’s home, while properties are manually encoded to reflect the unique requirements of each participant. The main contributions of this work are as follows: 1. A comprehensive framework integrating formal model verification with commercial sensors to capture ADL data in the homes of OAs. This is then extended with explicit contextual data. This facilitates the creation of personalised models and property specifications, linking ADL sensor events to OA preferences and needs. 2. A rigorous and systematic approach for the specification of models and properties based on information acquired from semi-structured interviews with each participant 3. Our approach is generalisable and we have applied it to a diverse range of OAs. Here we illustrate with two participants with contrasting ADLs. This paper is organised as follows: Section 2 discusses related work in sensor-based ADL monitoring and formal methods. Section 3 describes the main components of the framework and data collection. Section 4 describes the data processing and personalised models. Section 5 describes the experiments and the results. In Section 6 we present our conclusions and outline future work. 2 Related Work Early work on ADL monitoring using unobtrusive sensors has primarily focused on recognising and tracking ADLs. Initial studies, such as those by Munguia et al. [ 11 , 12 ] , used contact sensors and energy switches to capture events involving people living alone to track daily activities. Following this, a range of studies have examined various aspects of ADL monitoring. For instance, Kang et al. [ 13 ] focused on the use of IoT devices and sensor fusion to capture ADL events and improve data reliability, while Matsui et al. [ 14 ] proposed semi-automated mechanisms that relied on physical buttons for activity annotation. Tonkin et al. [ 15 ] worked on a multi-sensor dataset collected in a controlled residential setting, enhancing its accessibility through improved documentation, annotations, and organisation. That dataset comprises short-duration scripted ADLs. The use of formal methods in ADL monitoring has also been explored in previous studies. Konios et al. [ 16 ] used Petri nets to distinguish between normal and abnormal behaviour in data captured from motion, light, temperature, and contact sensors. Garcia-Constantino et al. [ 17 , 18 ] collected data from ambient and wearable sensors to identify abnormal behaviour during preparation of beverages in the kitchen using Petri nets. Magherini et al. [ 19 ] used temporal logic to automate the recognition of three ADLs within a hypothetical scenario involving a person living alone. The activities assumed specific sequences and timing of actions, which were captured by camera sensors. A shared characteristic of the above approaches is their reliance on common assumptions about the ADLs

Read the complete text