Mobile planner for proactive service composition

Abstract

Smart phone require an intelligent technology borne by their users. The users feel burdened by the interactions among many services on a small screen. We propose an AI (Artificial Intelligence) Mobile Planner for proactive service composition (i.e., a plan); it is a new planner for mobile environments that is equipped with a domain independent engine based on Hierarchical Task Network (HTN) planning that considers both contexts and user preferences. When the user's natural speech query (e.g., Best stew restaurant for a gathering of coworkers near Gangnam Station?) is received, the planner produces problem for plan based on predicate formula that consist of queries, contexts, preferences, and tasks. The context (i.e., location, weather, season, time, and so on) is inferred from sensors that are embedded in the smart phone. The planner dynamically recognizes the user's surroundings from the contexts, and considers the user's preferences based upon the user profile and history. It then selects and composes services that satisfy the problem, and executes web services based on cloud computing. We separate the description of the service model (i.e., domain (methods and operators) for plan) for a particular application from the general-purpose program used for service. An application designer writes services as user-friendly rule form in script file. After the planner read the file, he executes matched rules by new facts (problem). Thus, a non-technical designer only has to describe the service model in the domain. The validity of the planner's engine is experimentally confirmed by applying it to applications in various service domains: restaurant reservations, buying baseball tickets, and booking golf course. We are experimented using the data of the actual 200 testers for each domain and obtained 3 sample queries per user. The results about each domain showed the precision of 87.5%, 89.1%, and 92%, respectively. © 2012 American Scientific Publishers.