Strategic logics : complexity, completeness and expressivity

by transferring normative attributes from an agent to another. Such interactions are called delegation. Formal models of delegation and control were studied in, e.g., [189, 149, 191]. In this work, we consider the scenario where agents delegate control over propositions to other agents. The distinction between controllable and uncontrollable propositions stems from areas like discrete event systems and control theory, where, e.g., Boutilier [39] studied control in the context of deontic logic. Control and controllable propositions were also studied in [52, 66, 249, 248]. We now give an overview of the thesis. The main purpose of Chapter 2 is to introduce basic concepts and notation and to review relevant literature. The first section presents a brief survey on modal logic. Then, in sections 2.2, 2.3 and 2.4, we introduce epistemic, temporal and strategic modal logics and state known results that characterise their expressivity and computational complexity. In particular, we consider variants of ATL as extensions of branching-time logics. With such ATL-like logics we can describe dynamic multi-agent interactions. In Section 2.5, we discuss extensions of ATL with epistemic notions. Additionally, we suggest a framework for memory-bounded strategic reasoning. In particular, we introduce an epistemic variant of ATL that accounts for agents with limited memory resources as this case was neglected in the literature to date. In Chapter 3, we investigate the computational complexity of ATL and its epistemic extension ATEL. We show in detail how 'the complexity of the satisfiability problem for both logics can be settled at ExpTIME-complete. The part of the chapter about ATL is based on the paper 'ATL Satisfiability is Indeed ExpTIME-COmplete' by Walther, Lutz, Wolter and Wooldridge in the Journal of Logic and Computation, 2006 (265)' and the part about ATEL is based on the paper 'ATEL with Common and Distributed Knowledge is ExpTime-Complete' by Walther which was presented at the 4th Workshop on Methods for Modalities, Humbolt University, Berlin, December 1-2, 2005 [264]. In Chapter 4, we aim to extend the expressiveness of ATL without increasing its computational complexity. We introduce explicit names for strategies in the object language and extend modal operators with the possibility to bind agents to strategy names. In this way, we can fix the decisions of agents that possibly belong to several coalitions. By identifying the behaviqur of agents, we can reason about the effects of agents changing coalitions. Dynamic coalitions provide more flexibility to adapt abilities to a changing environment. We investigate the expressivity of the resulting logic ATLES and compare it to ATL and ATL*. Moreover, we formulate two model checking problems for ATLES and investigate their complexity as well as the complexity of the satisfiability problem for ATLES. Additionally, we present a complete axiomatisation. This chapter is based on the paper 'Alternating-time Temporal Logic with Explicit Strategies' by Walther, van der Hoek and Wooldridge which is going to presented at the 11th Conference on Theoretical Aspects of Rationality and Knowledge (TARK), Brussels, Belgium, June 25-27, 2007 [266].EThOS - Electronic Theses Online ServiceGBUnited Kingdo