Path planning is one of the critical technologies for robots to achieve autonomous operation, enabling to quickly find the safe path in complex dynamic environments. However, relying on global path planning alone cannot avoid dynamic obstacles, while only using local path planning may lead to falling into local minima and fail to reach the target. Therefore, a two-layer mobile robot navigation method suitable for complex and dynamic environments is proposed. In the first layer, a parallel sampling and bidirectional guidance rapidly-exploring random tree algorithm (PB-RRT) is employed to search the global path. This layer introduces a parallel heuristic sampling strategy and designs an evaluation function incorporating distance and angle factors to select the optimal sampling nodes for adaptive expansion. A bidirectional guidance mechanism is established to accelerate the merging process of the two trees. Then, combined with trajectory optimization methods, the initial path length and smoothness are optimized to obtain the key nodes of the path. In the second layer, the key nodes of the path obtained from the first layer are used as dynamic subtargets for the safe dynamic window approach (SDWA), which introduces an offset guidance method to avoid dynamic obstacles. Extensive experimental results indicate that the proposed method can effectively plan an optimal global path compared with other baseline algorithms. The ability to efficiently avoid dynamic obstacles in a dynamic environment verifies the feasibility of the proposed method.