Parallel Iterative Solution Methods for Markov Decision Processes

Applications of Markov decision processes arise in many problem areas including maintenance and replacement, inventory control and queuing models. However the extent to which Markov decision processes are used in practice is limited because the solution of realistic problems on serial machines is often impractical due to their large memory and processing time requirements. This talk will consider the use of parallel processors in the solution of Markov decision processes — an approach which has long been considered a potential solution to the computational intractability of these problems on serial machines.