Wandering and getting lost: the architecture of an app activating local communities on dementia issues

TL;DR

This paper presents the architecture and implementation of a scalable backend system for an app that helps persons with dementia avoid getting lost by real-time location tracking and anomaly detection, involving caregivers and volunteers.

Contribution

It introduces a microservices and serverless architecture for real-time dementia monitoring, demonstrating scalability and performance on a cloud platform.

Findings

Achieved efficient real-time data processing for dementia care

Demonstrated scalability with simulated load tests

Supported anomaly detection for wandering behavior

Abstract

We describe the architecture of Sammen Om Demens (SOD), an application for portable devices aiming at helping persons with dementia when wandering and getting lost through the involvement of caregivers, family members, and ordinary citizens who volunteer. To enable the real-time detection of a person with dementia that has lost orientation, we transfer location data at high frequency from a frontend on the smartphone of a person with dementia to a backend system. The backend system must be able to cope with the high throughput data and carry out possibly heavy computations for the detection of anomalous behavior via artificial intelligence techniques. This sets certain performance and architectural requirements on the design of the backend. In the paper, we discuss our design and implementation choices for the backend of SOD that involve microservices and serverless services to achieve efficiency and scalability. We give evidence of the achieved goals by deploying the SOD backend on a public cloud and measuring the performance on simulated load tests.