Uncertain decision making and modeling in multi-agent competitive environments.

摘要

An agent's decision-making in competitive multi-agent environments involves uncertainty and risk because other agents' knowledge and even goals may not be revealed. An agent has to make its decision about its further actions based on the uncertain information about the state of nature. In this thesis, an agent's decision-making process in competitive environments has been formalized, decision-making mechanisms have been analyzed to cope with different agents' interaction scenarios, a decision-making library named Agent-DMSL consists of five modules: rule-based module, Dempster-Shafer module, certainty module, risk module and uncertainty module has been built. This could fit into the big picture of agent-related software engineering technologies and make contributions to the design and implementation of software agent and multi-agent systems.