This paper intends to address the challenges of mobile intelligent agents in complex three-dimensional environments, specifically the difficulties in obtaining a globally feasible path and avoiding dynamic obstacles. For this purpose, a hybrid global dynamic path planning approach is proposed, which incorporates a neural network based six-direction search (SDS) algorithm and an improved dynamic window approach (DWA). Initially, the SDS algorithm is utilized to rapidly plan an initial feasible path. Subsequently, key nodes from this feasible path are extracted as sub-goal points for the DWA algorithm, ensuring that the path achieves global feasibility. When encountering dynamic obstacles, the mobile agent employs an improved dynamic window approach for local path planning. However, to enhance the environmental adaptability of the mobile intelligent, fuzzy logic is incorporated to dynamically adjust the weight of each parameter in the evaluation function. This combined approach enables the mobile agent to plan a globally feasible path in real-time within complex three-dimensional environments. Finally, both simulations and robotic experiments demonstrate that the proposed algorithm consistently enables mobile end-effector of the robotic arm to find a globally feasible obstacle-avoiding paths in complex three-dimensional spaces.