Font Size: 

Markov Decision Processes (MDPs) are often used in Artificial Intelligence (AI) to solve probabilistic sequential decision-making problems. In the last decades, many probabilistic planning algorithms have been developed to solve MDPs. However, the lack of standardized benchmarks makes it difficult to compare the performance of these algorithms in different contexts. In this paper, we identify important topological properties of MDPs that can make a significant impact on the relative performance of probabilistic planning algorithms. We also propose a new approach to generate synthetic MDP domains having different topological properties. This approach relies on the connection between MDPs and graphs and allows every graph generation technique to be used to generate synthetic MDP domains.

Markov Decision Process, Probabilistic Planning, Synthetic Domains Generation, Topological Diversity, Benchmarking