| This dissertation presents a distributed control model built on a decision-theoretic framework, where intelligent beings (i.e., agents with their own controllers) are capable of utilizing their resources and functions to perform tasks in a dynamically and probabilistically changing environment. The intellectual merit of this work is to model the decision making process of agents in consideration of the non-determinism caused by: (i) probabilistic agent's behavior, i.e., probabilistic outcomes for agent's actions due to failures, faults or other natural reasons; (ii) partially unknown and probabilistic environment, due to distributed nature of the system, lack of means for full observation within an agent or within the environment, lack of full knowledge about other agents' behaviors, and due to failures and faults occurring within the environment. The main novelty of this work comes from the way the decision making process of each agent, while modulated and distributed, constantly takes the benefit of what it learns from its environment and from itself. The decision making process of each agent includes synthesizing task plans, and executing these plans. |
