A Generic Approach to the Evolution of Interaction in Ubiquitous and Context-Aware Systems

This paper presents a model-based approach to the prob- lem of evolutionary adaptation of ubiquitous and context aware systems where it is di-cult or impossible to predict in advance the resources avail- able, the criteria for judging the success of the change and the degree to which human judgement must be involved in evaluation process. The model is introduced via a simple example based around the adaptation of web content to difierent output device conflgurations. The relationship of the model to software architectures is illustrated via an example imple- mentation in a homecare software framework. We argue that the model ofiers greater ∞exibility and control than other current approaches. The work presented in this paper concerns the adaptation of interactive systems where it is di-cult or impossible to predict in advance (i) the resources available, (ii) the criteria for judging the success of the change and (iii) the degree to which human judgment must be involved in the evaluation process. This is particularly of relevance in ubiquitous and context aware systems which must respond to a wide range of factors. Additional di-culties are expe- rienced when new devices or evaluation techniques are added and it is necessary to evolve the system using these new techniques. These challenges have been recognised by Dourish(1) and MacLean(2) who argue that continual adaptation of interactive systems should be regarded as the norm rather than the exception. Work in this area is aimed at addressing evolving systems such as ubiquitous systems within a home environment. The need for constant reconflguration of devices within a home environment has been discussed by O'Brien and Rodden(3, 4) who recognise that items within the home as well as the home itself are subject to continuous redesign. This evolution within the home environment has a knock- on efiect on interactive systems operating in that space - requiring them to be capable of evolving to deal with the new or changing requirements. The work presented here is a model-based approach to the dynamic conflgu- ration of interactive systems, based on the concept of generic and specialisable evaluation functions. These evaluation functions are responsible for analysing the options available for evolution to determine the optimal choice. Evaluation