Multilevel edge-disjoint fiber-tree planning and spectrum assignment for filterless optical networks with hybrid spectrum-cost objectives

Filterless optical networks (FONs) reduce node complexity and energy consumption through passive broadcast-and-select transmission, but their planning remains challenging because fiber-tree construction, demand assignment, and spectrum reservation are tightly coupled under tree-wide reservation constraints. This work studies multilevel edge-disjoint fiber-tree planning for FONs with spectrally disjoint levels and a 1500-km tree-diameter bound. A hybrid objective combining total spectrum width and equipment cost is formulated, and two solution frameworks are developed: a deterministic Layerwise Construct-and-Backtrack Refinement (LCBR) heuristic and a multiple-initial-state simulated annealing (MISSA) framework with feasibility-preserving neighborhood moves. Extensive simulations are conducted on five topologies, including standard benchmark networks, under fixed and varying traffic settings. The results show that MISSA generally achieves the best performance for the spectrum-dominant objective and remains highly competitive under hybrid spectrum-cost objectives among the compared methods. Runtime and statistical analyses further confirm that MISSA provides significant spectrum reductions in medium and large topologies, at the expense of higher offline computational cost. Under the equipment-only objective, however, the benefit of multilevel planning is limited, and single-level or small- K designs may remain competitive or preferable.