文件名称：Decision Making in Multiagent Settings
文件大小：1.7MB
文件格式：PDF
更新时间：2021-05-26 00:51:49
Decision Making Multiagent
overstated; every single day we are surrounded by more devices equipped with on-board computation capabilities taking care of the ever-expanding range of functions they perform for us. Moreover, the decreasing cost and increasing sophistication of hardware and software opens up the possibility of deploying a large number of devices or systems to solve real-world problems. Each one of these systems (e.g., computer, router, robot, person) can be thought of as an agent which receives information and makes decisions about how to act in the world. As the number and sophistication of these agents increase, controlling them in such a way that they consider and cooperate with each other becomes critical. In many of these multiagent systems (MASs), cooperation is made more difficult by the fact that the environment is unpredictable and the information available about the world and other agents (through sensors and communication channels) is noisy and imperfect. Developing agent controllers by hand becomes very difficult in these complex domains, so automated methods for generating solutions from a domain specification are needed. In this book, we describe a formal framework, called the decentralized partially observable Markov decision process (Dec-POMDP), that can be used for decision making for a team of cooperative agents. Solutions to Dec-POMDPs optimize the behavior of the agents while considering the uncertainty related to the environment and other agents. As discussed below, the Dec-POMDP model is very general and applies to a wide range of applications