Add pomp-value-iteration markdown.

Author: ctjoreilly

README.md

@@ -70,6 +70,7 @@ The file [aima3e-algorithms.pdf](https://github.com/aimacode/pseudocode/blob/mas
 |<hr/>|<hr/>|<hr/>|
 | &bull; |  | [VALUE-ITERATION](md/Value-Iteration.md) |
 | &bull; |  |  |
+| &bull; |  | [POMDP-VALUE-ITERATION](md/POMDP-Value-Iteration.md) |
 |<hr/>|<hr/>|<hr/>|
 | &bull; |  |  |
 |<hr/>|<hr/>|<hr/>|

md/POMDP-Value-Iteration.md

@@ -0,0 +1,20 @@
+# POMDP-VALUE-ITERATION
+
+## AIMA3e
+__function__ POMDP-VALUE-ITERATION(_pomdp_, _ε_) __returns__ a utility function  
+ __inputs__: _pomdp_, a POMDP with states _S_, actions _A_(_s_), transition model _P_(_s′_ | _s_, _a_),  
+      sensor model _P_(_e_ | _s_), rewards _R_(_s_), discount _γ_  
+     _ε_, the maximum error allowed in the utility of any state  
+&emsp;__local variables__: _U_, _U&prime;_, sets of plans _p_ with associated utility vectors _&alpha;<sub>p</sub>_  
+
+&emsp;_U&prime;_ &larr; a set containing just the empty plan \[\], with _&alpha;<sub>\[\]</sub>_(_s_) = _R_(_s_)  
+&emsp;__repeat__  
+&emsp;&emsp;&emsp;_U_ &larr; _U&prime;_  
+&emsp;&emsp;&emsp;_U&prime;_ &larr; the set of all plans consisting of an action and, for each possible next percept,  
+&emsp;&emsp;&emsp;&emsp;&emsp;a plan in _U_ with utility vectors computed according to Equation(__??__)  
+&emsp;&emsp;&emsp;_U&prime;_ &larr; REMOVE\-DOMINATED\-PLANS(_U&prime;_)  
+&emsp;__until__ MAX\-DIFFERENCE(_U_, _U&prime;_) &lt; _&epsi;_(1 &minus; _&gamma;_) &frasl; _&gamma;_  
+&emsp;__return__ _U_  
+
+---
+__Figure ??__ A high\-level sketch of the value iteration algorithm for POMDPs. The REMOVE\-DOMINATED\-PLANS step and MAX\-DIFFERENCE test are typically implemented as linear programs.