PaCR: An Agentic Preference-Aligned TAP Conflict Handling

With the increasing complexity of smart home environments, trigger-action-program (TAP) conflicts have emerged as critical issues, occurring when multiple automation apps compete for control over shared devices or interact in unintended ways. While existing approaches detect and resolve such conflicts using predefined rules or expert knowledge, they fail to account for the significant variability in individual pReferences for conflict resolution and significantly degrade the automation-system utility. To address this gap, we propose a preference-aligned conflict resolution (PaCR), a novel agentic framework that resolves conflicts while aligning with users’ personalized preferences for automation behavior. PaCR employs a two-step conflict resolution process: first, it interactively gathers contextual information from smart home configurations to generate conflict resolution agents (CRAs) with user profiles that accurately depict user preferences and characteristics; then, the agents, acting as users, utilize the semantic reasoning capabilities of large language models (LLMs) to assist in generating personalized conflict detection and resolution strategies. Due to the agentic conflict-resolving feature, we evaluate the effectiveness of PaCR with existing datasets at scale. PaCR demonstrates strong capabilities in addressing conflicts based on users’ needs and significantly improves conflict resolution efficacy and interpretability, offering a user-centric approach to managing smart-home automation conflicts.