Automated selection of the active device in interactive multi-device smart spaces

Pervasive computing offers the vision of seamless interaction by users with pervasive smart spaces filled with many wireless and wired devices, including monitors, speakers, printers, kiosks, soda machines, appliances, toys, sensors, and networked software services. To achieve seamless interaction, we must first address what we term as the active device resolution problem: when a user enters a smart room filled with N devices, which of the N devices is the “active device” that the user intends to interact with immediately? This paper focuses on the particular interaction mode introduced by remote control PDA’s and cell phones. Manual selection of the intended device, either by pointing-and-clicking, gesturing or selecting a device from a software menu, is either imprecise or decidedly intrusive. Automated selection of the active device, using context clues and user history, is a step towards the vision of more seamless interaction. In this paper, we study the historical behavior of a small sample of users and apply various prediction algorithms to user history to automatically select the next active device that a user intends to interact with. We show that the accuracy of prediction can vary depending upon the algorithm, e.g. up to 90% using 3rd order Markov prediction, and upon the length of training.