Autonomic ubiquitous computing: a home environment management system
Carlos Manuel Rodrigues Machado · Portuguese National Funding Agency for Science, Research and Technology (RCAAP Project by FCT) · 2010
The Ubiquitous Computing and Autonomic Computing reached a point of convergence in which pervasive technology in the environment meets the ability of people to interact with it, making use of all the possibilities made available by this technology. Ubiquitous computing envisions a habitat where the abundance of devices, services and applications allows the physical and virtual worlds to become seamlessly merged. The promise of ubiquitous computing environments is not feasible unless these systems can effectively "disappear". In order to achieve this goal, they need to become autonomic, by managing their own evolution and configuration with minimal user intervention. It is in this context that aspects like self-configuration and self-healing from the autonomic computing concept were adopted in this project. The context awareness and the creation of applications which use that context are the core concern of Ubiquitous Computing Systems and represent the fundamentals for autonomic actions in this type of systems. Such research raises questions on context acquisition, distribution and manipulation, as well as on artificial intelligence algorithms that decide autonomic actions in the environment, having implications in the human interaction with Autonomic Ubiquitous Systems. The research presented in this thesis concentrates on some of those issues. During this project it was developed an experimental setup for context acquisition, in an effortless way, of some activities of a small group of users. This experimental setup was installed in a real home where a young family, a couple and a small child, were actually living. This experimental setup was mainly responsible for the control of the light system of the house, by a network of several inter-connected devices scattered in the home with limited resources. This prototype installation allowed the validation of the system ability, to capture daily life behaviour patterns of the inhabitants. The system architecture was designed based on the concept of a high level and a low level autonomic management system taking from nature the model of the human reflex arc. A reflexive behaviour is managed at a local level by the small devices, with limited resources, high level management is responsible for processing and analysis of the events broadcast by the group of small devices, and run in a centralized mode in a PC. The concept of device information broadcast, to the communication medium, as events was used as an approach to: inter-connect future systems, monitor correct operation of the system devices, capture raw data for estimation of context; allow the visualization of system feedback in user interface devices. Finally, an algorithm using artificial neural networks in combination with simple statistics was developed which allowed the house to learn the routines of its inhabitants, making it truly intelligent by embedding the knowledge about patterns of activities of the users in the devices scattered in the environment, increasing their comfort and, at same time, leading to more energy efficiency. The analysis of the data captured, during two complete years, shows that the reduction of power consumption could be as high as 50%, depending on the profile of the usage of the light.