Upward book embeddings of DAGs: constraint-based methods and embeddability analysis
Abstract
The k-page upward book embedding (kUBE) problem is a fundamental challenge
in graph theory with applications in circuit layout, scheduling, and hierarchical visualization.
Despite its relevance, the problem—particularly for k ≥ 2—remains
underexplored. This thesis develops practical methods for solving kUBE and conducts
a detailed investigation of how graph structural properties influence upward
embeddability.
We first propose a Boolean satisfiability (SAT) encoding, SAT-1, that extends
existing k-page book embedding techniques to the general kUBE setting. For the
special case of k = 2 (2UBE), we introduce SAT-2, a more compact SAT encoding
exploiting the fixed number of pages, and a constraint programming (CP) model as
an alternative formulation. Empirical evaluation shows that SAT solvers consistently
outperform CP, with SAT-2 achieving up to 40% faster runtimes on large instances
and up to 30× speedups on hard instances from the North dataset compared to
SAT-1.
Beyond solving efficiency, we systematically analyze how upward book embeddability
depends on structural parameters such as the edge-to-vertex ratio (m/n).
Through exhaustive enumeration and sampling, we identify sharp phase transition
phenomena across different values of k (up to k = 6) and model the phase transition
threshold as a function of graph size and page count using a power-law relationship,
providing the first quantitative characterization of this phenomenon.