Dynamic Decision Problems with Cooperative and Strategic Agents and Asymmetric Information

Abstract: There exist many real world situations involving multiple decision makers with asymmetric information, such as in communication systems, social networks, economic markets and many others. In this dissertation, we attempt to enhance the conceptual understanding of such systems and provide analytical tools to characterize the optimum or equilibrium behavior.

Specifically, we study four discrete time, decentralized decision problems in stochastic dynamical systems with cooperative and strategic agents. (i) We study energy-delay trade-off in a relay channel through stochastic control techniques for cooperative agents with decentralized information. (ii) Extending this model for strategic users, we study a general model dynamic game with asymmetric information and independent types where we present a forward/backward sequential decomposition algorithm to find a class of perfect Bayesian equilibria (PBE) of the game. (iii) Using this methodology, we study a general two player dynamic LQG game with asymmetric information, where we show that under certain conditions, players' strategies that are linear in their private types, together with Gaussian beliefs, form a PBE of the game. (iv) Finally, we consider two sub problems in decentralized Bayesian learning in dynamic games. (a) We consider an ergodic version of a sequential buyers game where we design incentives to align players' individual objectives with the team objective. (b) We present a framework to study learning dynamics and especially informational cascades for general decentralized dynamic games with independent types and noisy observations, where we characterize informational cascades for a specific learning model.