Association rule-based Markov Decision Processes

In this paper we present a novel approach for the fast solution of Markov decision processes based on the concept of association rules. These processes have successfully solved many probabilistic problems such as: process control, decision analysis and economy. But for problems with continuous or high dimensionality domains, high computational complexity arises, because the search space grows exponentially with the number of variables. In order to reduce the complexity in these processes, we propose a new approach to represent, learn and apply the actions that really operate on the current state as a small set of association rules and a new value iteration algorithm based on association rules. Experimental results on a robot path planning task indicate that the solution time and therefore the complexity of the proposed approach are considerably reduced, because they increase linearly when the number of the states increases.