Personalized adaptation in pervasive systems via non-functional requirements

Abstract

Pervasive environments are socio-technical systems that support the daily routines of their users in an invisible and unobtrusive manner. These systems are aware of and adapt to both the operational context and the characteristics and preferences of their users. Designing adaptation mechanisms that guarantee maximal user satisfaction is challenging, due to the inherent differences between users and the changing context where the system operates. In order to tackle this problem, we propose an approach that compares alternative system behaviors in terms of how well they satisfy the preferences of the current user concerning Non-Functional Requirements (NFRs) such as efficiency, comfort, energy saving, etc. Specifically, we propose a model-driven framework in which the models represent the user routines that the pervasive system helps to achieve. These routines include variability points, thereby enabling their behavior to be adapted at runtime in order to fit the context and the user preferences over NFRs. Our contributions include: (i) user-adaptive task models, a modeling language to describe user routines that accounts for user preferences over NFRs; (ii) algorithms that use our models at runtime to guide a pervasive system in adapting its behavior to user preferences and context; and (iii) an implementation and evaluation of our techniques.

A procedural and flexible approach for specification, modeling, definition, and analysis for self‐adaptive systems

“…These techniques would also facilitate the construction of initial reference values to assist in the interpretation of results, which is pointed out as another challenge. [7][8][9] However, these possible solution paths lead us to another challenge, pointed out by the same authors: the complexity of a self-adaptive system must be considered in simulations and there is a difficulty in reproducing these conditions as they exist in real-world application scenarios. 10,11 Thus, research paths that strive to advance in the areas of performance analysis of self-adaptive systems at design time are fundamental in an attempt to overcome problems still present in the area.…”

Data-Driven Supervision of Autonomous Systems

“…Serral et al [249], propose run-time algorithms for the selection of optimal tasks when adaptation is triggered by changes in user preferences, by faults in the execution of some tasks, by plan failures or by context evolution. In order to select the best set of subtasks, they evaluate the contribution of the possibilities to the NFRs.…”