Abstract: To address insufficient turning radius, excessive angle deviation, and unstable boundary connections between partitions in offset paths during automated fiber placement of carbon fiber reinforced resin matrix composites on complex surfaces, a partitioning method based on constraint triggering and genetic algorithm is proposed. The initial seed points are determined through curvature screening, farthest point sampling, and coverage-based selection. The turning radius and angle deviation are introduced as placement feasibility constraints to identify partition boundaries during path offsetting. The selection of seed points for new partitions is transformed into a one-dimensional parametric optimization problem along existing boundary curves, and the seed point positions are optimized by combining genetic algorithm search with offset distance optimization. Furthermore, a calculation method for evaluating the coverage rate, overlap rate, and gap rate of carbon fiber prepreg tows is established based on triangular mesh surface sampling, enabling quantitative assessment of partition placement quality.A wing surface model and a half-inlet surface model are employed for verification. The results demonstrate that the proposed method can generate continuous and stable partitioning paths under different ply angles. Compared with the sector-based partitioning method, the average number of partitions is reduced from 8.63 to 3.88, the average coverage rate is increased to 99.15%, and the average overlap rate and gap rate are reduced to 0.36% and 0.49%, respectively. The proposed method effectively reduces the number of partitions and decreases overlap and gap defects, thereby improving the partitioning quality of automated fiber placement on complex surfaces.