A Kinematically Constrained Segment Library for Multi-Target Multirotor Inspection Route Planning in 3D Mountainous Terrain

Mountainous inspection couples visit ordering with a feasible three-dimensional connection between consecutive nodes. A return-to-base (RTB) rule makes the last visit consequential because it determines the closing segment. We address these decisions in two stages. We first build a segment library for all candidate directed transitions among the base and prescribed inspection nodes. The builder enforces terrain clearance, altitude, climb and descent, and within-segment turn constraints. Stored transitions then form M-leg open (OPEN) routes or (M+1)-leg RTB routes. A random-key improved artificial lemming algorithm (IALA) searches visit orders. Before a candidate consumes a counted evaluation, its permutation-order guard (POG) checks for a parent-identical or previously evaluated order and tries an insertion move. Independent recomputation verified 7,214 stored segments; a separate audit reconstructed 800 routes without finding an unavailable transition in any selected route. Among six generic metaheuristics, IALA ranked first in all four OPEN scenarios, with mean ranks of 1.25 to 1.40, and all 20 prespecified contrasts remained significant after Holm adjustment. ACO led the route-specific OPEN comparison. Removing POG caused the clearest ablation loss (Holm-adjusted p = 2.18e-11; A12 = 0.975). All paired S01 and S04 RTB runs met the implemented model-feasibility criterion. In the real-DEM S04 simulation, RTB raised the median paired path length by 2,225.5 m. These results concern synthetic cases and one real-DEM planning simulation; they do not establish flight readiness.