Hierarchical Task Network (HTN) planning has attracted much attention recently, yet its support for numeric reasoning is still very limited. This paper focuses on totally‑ordered numeric HTN (TOHTN) planning and shows how a standard SAT‑based encoding can be naturally extended with SMT to handle numeric fluents. The approach augments each Boolean clause with SMT constraints such as $x \ge 5$ for numeric preconditions or equality/inequality expressions for effects, resulting in a hybrid SAT‑SMT formulation that retains the compactness of SAT while gaining numeric reasoning capabilities.
To foster experimental research, we introduce a benchmark suite for numeric TOHTN planning, covering a range of task networks with varying sizes and numeric complexities. The suite provides a common evaluation platform. Experiments compare pure SAT encodings, the proposed SAT‑SMT encoding, and several existing numeric planners. Results indicate that even this simple SMT extension yields a competitive baseline, especially on instances with many numeric constraints.
This work opens new avenues for expressive numeric HTN planning. Future extensions may incorporate richer numeric constraints, optimization objectives, or learning‑driven task decomposition to improve both expressiveness and solving performance.
Review