Intelligent Systems Lab Amsterdam
Please see the new ISLA website at: http://isla.science.uva.nl
Planning under uncertaintyPlanning is the process of computing a sequence of actions that fulfill a given task as well as possible. It is a crucial part of any intelligent agent; human, robot or software agent alike. We are interested in planning under various forms of uncertainty. First of all, the agent might be uncertain regarding the exact consequence executing a particular action. Furthermore, the observations the agent receives through its sensors might be noisy or provide only a limited view of the environment. When the agent is part of a team, a third source of uncertainty are its teammates, as it should consider their actions as part of its own decision making. We study scenarios in which agents have the ability to communicate, but bandwidth is limited and communication has a certain cost.Research group membersdrs. M.T.J. Spaandr. N. Vlassis Research achievementsPartially observable Markov decision processes (POMDPs) provide a rich mathematical framework for acting optimally in partially observable and stochastic environments. However, computing optimal plans is intractable in general, and so we focus on approximate methods in order to able to solve more interesting problems. We have developed a successful approximate POMDP solver called Perseus for single agent planning problems, which exploits structure present in real-world planning domains. Extending our work to a multiagent setting we are developing approximate planning methods in which the agents optimize their information exchange.FundingThis research is partially supported by PROGRESS, the embedded systems research program of the Dutch organization for Scientific Research NWO, the Dutch Ministry of Economic Affairs and the Technology Foundation STW, project AES.5414.LinksFor more information, relevant publications, and software please check the webpage of Perseus. |
| Maintained by Bas Terwijn. Last edited on Mon, 25 Jan 2010 13:38:55 +0100 |