A Service-oriented device selection solution based on user satisfaction and device performance in a ubiquitous environment

Ubiquitous computing provides the vision of a smart space filled with various smart devices and services where users can navigate with seamless interaction. To achieve such an intelligent scene, service providers attempt to fulfill the functional requirements of users as well as their non-functional demands, such as user satisfaction. Such non-functional demands depend on required services as well as the devices that execute them. For instance, different devices may provide different user satisfaction for an identical service. Currently, selecting a desired device for a requested service is generally completed manually. However, it is a challenge to match a desired device to a service in the aforementioned smart environment because a service is an abstract concept associated with user intention and experience, whereas a device is a concrete item associated with physical functionalities. Therefore, an automatic mechanism is imperative to balance the benefits and drawbacks of selecting a specific device to instantiate a particular service for a certain consumer. The objective of the current paper is to propose a solution that automatically selects one device among multiple devices with overlapping or identical functionalities for a specific given service to achieve maximum user satisfaction. To achieve this, the device selection procedure in a ubiquitous environment is decomposed into four tasks: 1) filtering to obtain candidate devices that are functionally available for the given service, 2) predicting user satisfaction with these candidate devices considering historic usage, 3) estimating device performance based on available device resources, and 4) selecting the top-ranked device to achieve maximum user satisfaction. This study constructs a quantitative metric for selecting a device for a certain service by quantitating service-oriented device performance based on available device resources and predicting user satisfaction considering historic usage. This method is validated through a comparison to another quantitative solution that considers user preference and a method that simply considers device capabilities. Using an example scenario, the overall hit rate of this proposed method exceeds those of the other two methods, and the experimental results indicate that the proposed method is helpful for automatically selecting a user-desired device for a specific service.