Cellular Learning Automata for Competitive Loss Sharing

Abstract

In this chapter, we provide an overview of the problem of loss sharing with competitive service providers. The irregular form of the cellular learning automata (CLA) is used to model service providers and relationships among them. In competitive conditions, service providers try to reach a compatible loss-sharing agreement (LSA). Under a compatible LSA, no service provider can improve its utility by changing strategies unilaterally (e.g., the shares from the LSA) while the neighbor service providers keep their strategies unchanged. To offer an irregular cellular learning automaton (ICLA) based loss-sharing approach, we first extend the theory of ICLA by presenting a new local rule for convergence of ICLA to a compatible point. Formal proofs for the convergence are provided, and results of the conducted experiments support the theoretical findings. The convergence of ICLA to the compatible point is of great importance because it can conduce to efficient solutions for the problems and applications. As an application of the presented local rule, we present an ICLA-based loss-sharing approach for competitive service providers. We demonstrate that using ICLA, finding compatible LSAs do not need detailed information about distributions of the losses and even utility functions of the other service providers. The results of the conducted experiments illustrate that the proposed approach can improve the utility of service providers and users.