Enhancing a simulator and visualization platform for human activity datasets : PacSim

Abstract

Pervasive computing is constantly increasing in popularity. Widespread mobile technologies have placed immense computational power in the palms of our hands. Using these technologies, we can gather large amounts of human activity data which can drastically improve our lives, particularly at risk individuals such as dementia-care patients. Unfortunately, constant human activity data can be expensive and obtrusive to gather. Researchers instead opt to collect shorter spreads of data, using tools to generate the data in between. Thus, this work proposes the optimization of PacSim (Pervasive Activity Context Simulator). PacSim was originally a system proposed in a past dissertation which featured local server set up for human activity data collection as well as minimal simulation capabilities [1]. This work overhauls the simulation engine, making use of machine learning and pattern recognition to generate human activity data. Data is simulated in a central server, providing analytics and real-time evaluation to the client through a modern web platform. In addition, PacSim includes a visualization tool that allows researchers to visually analyze and perform run-throughs of human activity datasets. The tool allows researchers to create floorplans as well as place activity points and safety zones. The tool then provides a comprehensive report measuring the accessibility and safety of the floorplan. All of this is deployed to a web server as a modern web application using .NET as a backend and SQL Server as a database engine. Evaluation is carried out to determine accurate algorithms for human activity classification which is then implemented into the platform.