Automated planning is essential for many applications, such as autonomous systems, robotics, and intelligent decision support, where agents must make decisions and execute actions to achieve defined goals. However, as planning domains grow in scale and complexity, classical planning techniques sometimes struggle to generate effective plans. Hierarchical goal decomposition is an approach to classical planning that applies divide-and-conquer principles to break down planning problems into smaller, more manageable subproblems, allowing planning systems to reason about each subproblem incrementally. Even so, the classical planning paradigm assumes there is complete, deterministic information about the environment and the effects of actions—an assumption that often does not hold in dynamic, real-world scenarios.

This research will build upon two fundamental approaches for goal decomposition from classical planning and extend them to probabilistic planning: landmarks, which are intermediate conditions satisfied at some point in every solution plan; and hierarchical goal networks, which are methods to decompose goals into sequences of subgoals. Furthermore, we will investigate how probabilistic planning algorithms, like Monte Carlo Tree Search, can be adapted to effectively leverage information on goal decompositions from probabilistic landmarks and probabilistic hierarchical goal networks to improve planning performance in stochastic domains. By bridging the gap between hierarchical goal decomposition and probabilistic planning, this research will develop more scalable and adaptive planning techniques to address the growing complexity of planning domains, improving planning systems for real-world applications where scale and uncertainty pose significant challenges.