Preference Handling in Relational Query Languages Radim Nedbal Institute of Computer Science Academy of Sciences of the Czech Republic Pod Vodárenskou vezí 2, 182 07 Prague 8, Czech Republic Email: radned@seznam.cz Abstract—The need for handling preferences in relational query languages arises naturally in real-world applications that deal with possible choices generated by the current state of the world captured in the relational data model. To address this problem, we propose a fully declarative language for encoding preferences conditional on the current state of the world represented as a relation database instance. Being based purely on the qualitative type of information about the preference model, the language supports encoding intuitive statements about preferences and having constructs for various kinds of preferences, the language leads to a flexible approach for specifying the most desirable choices of autonomous systems that act on behalf of their designers. Throughout the paper, we use an example of a control support system for a bank surveillance to motivate the need for our framework and to illustrate it. Keywords-conflicting preferences; comparative preference statements; various kinds of preferences I. I NTRODUCTION The Problem and the Main Goal: Complex autonomous systems make many decisions during their run time in environment where they act. The decisions are driven by designer’s desires, and the environment is usually dynamic. Representing and processing desires in terms of preferences, which order possible choices (decision outcomes) so that a more desirable choice precedes a less desirable one, appears to be especially appealing as it allows one to specify desires in a declarative way and to deal with inconsis- tencies and exceptions in a quite flexible manner. Indeed, a preference-based approach is amenable to customization both by allowing specification of additional desires and providing additional information about the environment. As the relational data model (RDM) allows to describe a generic model of a dynamic environment, where the number and properties of concrete object instances can change, our main goal has been to develop a general framework for preference handling in relational query languages (RQL) to support the user-friendly design of autonomous systems that can act in a dynamic environment. To this end, we propose a fully declarative language for encoding preferences conditional on the current state of the world represented as a relation database (DB) instance and show that the most desirable choices w.r.t.designer’s preferences encoded in the language can be denoted as a relational query that takes the DB instance as input. • First, we give the syntax (Subsect. II-B on page 3) and present main ideas underlying the model-theoretic se- mantics (Subsubsect. II-C1 on page 3) of the language that supports very intuitive and compact specifying of the model of designer’s preferences in eight different, qualitative (both precise and imprecise) ways. • We outline a non-monotonic reasoning mechanism (Subsubsect. II-C2 on page 5) yielding the so-called distinguished preference models (DPMs) (Par. II-C2 on page 5) and show that the DPM semantics is both intuitive and well-defined (Par. II-C2 on page 5) for any set of formulae of the language, which are inherently in conflict with each other because of tradeoffs inherent in preferences they encode. • Finally, we briefly outline a constructive semantics (Subsect. II-D on page 5) that agrees with the declar- ative semantics and is based on special, compact representation (Subsubsect. II-D1 on page 5) of the information communicated by the language that allows to exploit disjunctive datalog machinery to identify the most desirable choices and to denote them in a RQL (Subsubsect. II-D2 on page 5). Motivation: System Configuration & Design: To see the need for our framework, consider a command and control room for real-time operations. According to [1], this can be a control room for handling large-scale emergencies such as earthquakes, flooding, terrorist attacks; a control center for complex missions such as NASA shuttle missions; an army command center; or a monitoring center for some company that receives a lot of real time data that requires the attention of a decision maker. To be concrete, let us focus on the setting of a bank surveillance system with infrared (IR) and non-infrared cameras located in various places. For the sake of illustration, see Fig. 1, depicting a room with a single gate, two light sources, four non-IR cameras A1, . . . , A4, and a single IR camera A5. In a realistic scenario, we are talking about tens of rooms and hundreds of information streams to which the decision maker in the control room might have access. Clearly, no decision maker can handle this volume of information. Therefore we aim at a feasible solution that can have 978-1-61284-832-7/11/$26.00 ©2011 IEEE