With the development of fields such as autonomous driving, drone control, and robot navigation, autonomous navigation and path planning in three-dimensional environments have become one of the key technologies. Traditional two-dimensional path planning methods cannot meet the navigation needs in complex three-dimensional environments. This paper proposes a three-dimensional environment navigation and path planning method based on deep reinforcement learning (DRL), which uses a combination of deep neural networks and reinforcement learning algorithms to achieve adaptive navigation of the agent in complex three-dimensional environments. By designing a continuous action space, optimizing the reward mechanism and combining a state representation with dynamic environmental features, the method in this paper effectively improves the obstacle avoidance performance of the agent in a multi-obstacle environment. Experimental results show that compared with traditional path planning methods, the DRL model in this paper has a higher success rate, more accurate path planning and stronger adaptability in complex 3D environments. This paper provides an efficient and robust solution to 3D navigation and path planning.