Simpele, Flexibele en Open-Source 2D Cutting & Packing Algoritmes

Cutting and packing (C&P) problems involve placing items into larger containers while optimizing objectives such as minimizing waste or maximizing material yield. They arise in numerous industrial and logistical contexts where solution quality directly affects operational efficiency.<br />However, their combinatorial nature makes them notoriously difficult to solve. The number of possible configurations grows exponentially with problem size, challenging the design of methods that can deliver high-quality solutions in a reasonable time. Despite decades of research, the quest for more effective algorithms remains pressing, with the pursuit of marginal gains often leading to increasingly fragile and convoluted methods that fail to generalize. This thesis advances the state of the art in 2D C&P problems and focuses on developing simple, flexible, open-source algorithms. It comprises three core contributions, each accompanied by a robust open-source implementation. First, we introduce GDRR, a goal-driven ruin and recreate heuristic for the 2D rectangular bin packing problem with guillotine constraints. We then focus on irregular C&P problems, where polygonal shapes add a significant geometric challenge on top of the underlying combinatorial challenges. To address this, we present jagua-rs, a dedicated 'Collision Detection Engine' that can efficiently and accurately verify whether items are colliding, thereby lowering the barrier to entry for 2D nesting. Finally, building on this engine, we introduce sparrow, a complete heuristic that sets a new academic and open-source baseline for 2D irregular strip packing problems. A lack of reproducibility hinders researchers from verifying or building on prior work, impeding scientific progress. Within the field of 'Computational Operations Research', reproducibility unfortunately remains the exception rather than the rule. We contend that a culture of openly sharing source code would largely mitigate this issue. This thesis takes a strong stance on reproducibility: the codebases of the implementations are treated as first-class research artifacts — on par with, and complementary to, the methodological descriptions. This dissertation puts forward the idea that pursuing simple, flexible methods — paired with polished open-source implementations — presents one of the most promising paths to lasting impact for the research community. In this way, we aim for our contributions not merely to be a single step forward, but to catalyze future research and advances.