3D path planning is critical for autonomous systems operating in complex environments. However, existing meta-heuristic algorithms face limitations in handling high-dimensional obstacles, computational inefficiency, and premature convergence. To address these challenges, this study proposes a hybrid framework integrating a periodic pattern search reward mechanism into active subspace particle swarm optimization algorithm, combined with a novel mutation mechanism to enhance population diversity, thereby balancing exploration and exploitation. Furthermore, by leveraging the geometric invariance of B-spline curves, a multi-phase optimization strategy is developed to iteratively refine sparse control points from coarse to fine resolutions, ensuring globally optimal and smooth 3D trajectories while mitigating computational complexity. Comparative evaluations against benchmark algorithms demonstrate that the proposed framework performs better than the existing ones in path optima, convergence speed, and robustness across diverse scenarios, including cluttered urban environments and undulating marine terrains.