CoPI: Enabling Probabilistic Conflict Prediction in Smart Space Through Context-awareness

In a smart space influenced by multiple parties, conflicts can arise when competing users try to control the same devices in different ways. Such conflicts usually require user negotiation to resolve and thus lower people's satisfaction and trust in the smart system. Finding a conflict is the first step to resolving it, and the timing when a conflict is identified impacts the options for resolution. Most existing approaches identify conflicts only at the time they occur, which offers little help to the users in resolving the conflicts, especially without them having to compromise. A better solution is to predict potential conflicts in advance so that the users can coordinate themselves to avoid conflict situations beforehand. In this paper, we propose a novel context-aware conflict prediction framework that addresses the research gaps identified in existing literature. We mine habit patterns from the user's previous interactions with smart devices in the various environments they occupy. These habits serve as inputs to our conflict prediction algorithm which takes the habits of pairs of users and outputs context situations in which those users have the potential to conflict. To support eventual flexible resolution, we use explicit models of the uncertainties of users' behaviors to associate each potential conflict scenario with a probability of that conflict occurring for these particular users. We evaluate our framework on real-world datasets to demonstrate the effectiveness of the proposed approach.