Partially observable Markov decision process

Overview

The Partially Observable Markov Decision Process (POMDP) is an extension of the Markov Decision Process. POMDPs are used for planning actions where the entire world, or state space is not known during planning. Currently, most POMDPs are computationally intractable, so computer scientists have developed "policies" to approximate the solutions to POMDPs.

POMDPs are often used in solving complex path planning problems for mobile robots.

References

Many papers have been written in the last five to ten years on POMDPs.

• N. Roy & S. Thrun. "Coastal Navigation with Mobile Robots". Advances in Neural Information Processing 12 (NIPS). Colorado, Dec. 1999 [1]