POMDP Partially observable Markov decision process

<p>Mar 27, 2012 08:14 AM

&#160;

genl of Markov Decn Proc.

&#160;

must maintain a prob distrib over the set of poss states;

based on a set of observ and observ probs and the underlying MDP.

&#160;

real world &#160;sequnt. decn proc... &#160;robo navigation; &#160;planning under uncertainty.

started in OR community &#160;move to AI and Automated Planning.

&#160;

<b>An exact solution to a POMDP yields the optimal action for each possible belief over the world states</b>...

<b>The optimal action maximizes (or minimizes) the expected rewar</b>d (or cost) of the agent over a possibly infinite horizon.&#160;

<b>The sequence of optimal actions is known as the optimal policy of the agent for interacting with its environment</b>.</p>