OPTIMIZATION ALGORITHMS FOR STABLE AND EFFICIENT BIN PACKING IN ROBOTIZED PALLETIZING STATIONS

The thesis work addresses the design of optimization algorithms for stable and efficient pallet loading in robotized logistics systems, considering both offline and online variants of the three-dimensional bin packing problem. After reviewing the literature on pallet loading and multidimensional packing, a 3D MILP formulation is proposed to generate multi-pallet loading plans that satisfy realistic constraints on geometry, load-bearing capacity, stability, orientations, and pallet height. On the online side, three controllers are developed and compared: an enhanced Deep Bottom-Left heuristic (DBL+), a beam-search palletizer (BEAM), and a novel stochastic model-predictive buffering planner (SMPBP), all operating under streaming arrivals of items with unknown future demand. Experimental analyses quantify the trade-offs between packing quality, computational latency, and algorithmic complexity, showing how buffering, limited look-ahead, and structured search can substantially improve online performance while remaining compatible with industrial cycle times. The results provide both methodological insights and practical tools for next-generation automated palletizing cells.