Assurances for Decentralized Self-Adaptive Systems: 2014-2017

Research project funded by Swedish Research Council (VR).

Software systems will increasingly be integrated with one another. One example domain is smart homes that employ sensors and actuators for providing care to elderly people or improving energy consumption. Engineering these systems and guaranteeing the required qualities (reliability, efficiency, personalization, etc.) during operation is complex due to uncertainties resulting from incomplete knowledge at design time. Examples of such uncertainties are changing user needs, subsystems that come and go at will, and faults that are difficult to predict. These challenges have motivated the need for self-adaptive software systems. A self-adaptive software system consists of a managed system and a feedback loop that monitors and adapts a managed system when needed to realize some quality goals. Despite substantial achievements in the field, important challenges remain open for future research. One key challenge is to assure the required quality properties in decentralized self-adaptive software systems. For a decentralized self-adaptive system, the software of the managed system is deployed on a network of computing devices, and this software is controlled by multiple feedback loops that act locally.

To tackle this challenge, we propose an innovative approach that combines formal modelling and verification with online learning. Formal underpinning of the design of self-adaptive systems enables automatic verification of required quality properties. To design feedback loops, we employ principles for control theory, which provide a mathematical basis for analysing key properties of self-adaptation in decentralized settings, including stability and transient behaviour. Online learning enables the self-adaptive system to acquire knowledge about design time uncertainties during operation. To steer and validate the research we use case studies in the domain of smart homes with a focus on self-configuration, context-aware adaptation, and self-healing. Self-configuration enables the system to dynamically discover and configure new services based on personal needs and conditions. E.g., the system may activate a new service that enables an elderly to alarm a welfare helper via voice when he/she enters the bad room at night. Context-aware adaptation enables the system to adapt with the context of use. E.g., a mobile application may offer a welfare helper a service that provides specific information once she approaches the home of an elderly. Self-healing allows the system to detect errors and autonomously recover. E.g., if a sensor for tracking falls of an elderly fails, the information may be inferred from data collected by other sensors. Online learning allows gradually improving and anticipating adaptations of services based on personal usage and changing operating conditions.

Push the Line! Reliable Assistive Technologies to Empower Elderly People to Live Independently Longer

Research project in collaboration with Göran Petersson of the eHealth Institute, funded by VINNOVA.

Project partners: Växjö Kommun, Mönsterås Kommun, Landstinget Kronoberg, Sigma IT, Wexnet, and Växjöhem.

The growing population of elderly people poses challenges to our society. It is generally acknowledged that innovation in Information and Communication Technologies (ICT) will play a key role in enhancing independence of elderly people, reduce costs for care, and offer a great business potential. Despite vast research, the adoption of these technologies has not taken off yet, mainly because a lack of reliability of the solutions. The Push the Line! project takes a holistic perspective and put reliability of assistive technologies central: from a user perspective (user-centric solutions), technological perspective (correctness of systems), and organizational perspective (sustainable arrangement of the stakeholders). The project focuses on elderly people living in their own houses that are become increasingly uncertain about whether they are able to cope with challenges of their daily life. The overall aim is to support these elderly people with reliable assistive services to deal with these challenges pushing the line for them to live independently longer and reduce costs for elderly care. Examples of such services are reminder functions, tele-monitoring, and support for remote rehabilitation. Large-scale validation and market potential are key aspects. For the first stage of the project, the team received 500 kkr from Vinnova to identify concrete needs and perform a market analysis, identify concrete goals and prioritize them, identify most promising research tracks, perform targeted pilot studies, and investigate current regulations and legal aspects.

A Foundation for Engineering Decentralized Self-Adaptive Software Systems: 2012-2015

Marie Currie Career Integration Grant funded by the EU's 7th Framework Programme (FP7).

Self-adaptation enables a system to adapt itself autonomously to changes to achieve particular quality goals. Despite substantial achievements in the field, major scientific and technical gaps must be addressed to deal with decentralized control of adaptation, which is required in a growing class of software systems such as networked smart homes, large-scale mobile applications, and multi-robotic systems. The overall goal of this project is to study and develop a foundation for engineering decentralized self-adaptive systems. Our focus is on the application of formal methods to provide assurances for decentralized self-adaptive systems, and the use of empirical methods to validate solutions. The research results will be applied to concrete applications in the domains of smart homes for elderly care and robotics. For more information and results, we refer to the project website.

CareSmart: Improving Welfare Services for Elderly People Using Smart Homes: 2012-2013

Project in collaboration with Nybro's Welfare Department (Nybro Wikipedia) and Linnaeus Information Engineering Center (Interest Group on Architectural Knowledge). An important role of a community's welfare department is providing welfare services to elderly people living in private homes. Such services are provided by trained welfare helpers and professional nurses that travel between the homes of the elderly people. In the CareSmart project, we focussed on the nightly visits of the elderly people. Traditionally, the ambulant personnel use predefined schedules for the visits. Alarms invoked by elderly people may require dynamic rescheduling. However, welfare helpers and nurses lack important data about the actual conditions of the elderly people, which results in inefficiencies of visits, accidents, and from time to time even catastrophes. The goal of this project was to perform a domain study for improving the welfare services by exploiting smart home technology; i.e., homes equipped with unobtrusive sensing and actuating infrastructure, intelligent devices, and other technology to support assisted living of the elderly. The main findings are a classification of needs in three groups: (1) unnecessary visits of welfare helpers that could be avoided (e.g., an elderly sleeps quietly), (2) anticipatory visits that could anticipate a lot of overhead, if detected in time (e.g., a saturating diaper), and (3) critical visits that could avoid severe problems (e.g., an elderly fell). Furthermore, the study revealed that smart homes technology needs to be viewed from the perspective of their embeddedness in the social and cultural practices of everyday living, we requires a multi-disciplinary approach for research. One of the publications with some additional information: MIT 2013.

E'SPLS: Software Product Lines for Logistic Systems: 2009-2011

R&D project in collaboration with Egemin. This project was funded by the Institute for the Promotion of Innovation through Science and Technology in Flanders (IWO). In this project we developed an architectural approach for online updates of deployed software product line products. The approach comprises of two complementary parts: (1) an update viewpoint that defines the conventions for constructing and using architecture views to deal with multiple update concerns; and (2) a supporting framework that provides an extensible infrastructure supporting integrators of a software product line. The approach was successfully applied and empirically validated to Egemin's software product line for logistic systems. Publications with additional information: WICSA 2011, and ESEM 2008.

EMC2: Egemin Modular Control Concept: 2004-2007 (Movie: 12M, 98M)

R&D project in collaboration with Egemin. This project was funded by the Institute for the Promotion of Innovation through Science and Technology in Flanders (IWO). In this project, we studied, designed, developed, and evaluated a decentralized control system for automatic guided vehicles. The driving motivations behind decentralizing the control of AGVs were new and future quality requirements such as flexibility (deal autonomously with dynamic operating conditions) and openness (deal autonomously with AGVs entering and leaving the system). Some publications with additional information: AAMAS 2005, JAAMAS 2008, and OOPSPLA 2008. The book Architecture-Based Design of Multi-Agent Systems provides an in dept description of the design and realization of the AGV transportation system. Here is a link to a movie that shows a prototype realization with the main features: 12M movie, 98M movie.