Autonomous navigation and control technologies for mobile robots in complex dynamic environments represent a significant research direction in the fields of artificial intelligence and robotics. However, traditional control algorithms exhibit certain limitations when faced with high-dimensional state spaces and uncertain environments. To enhance the path planning and obstacle avoidance capabilities of robots in complex environments, a control algorithm based on deep reinforcement learning, specifically the PPO-DQN algorithm integrating Proximal Policy Optimization (PPO) and Deep Q-Network (DQN), is studied. Experimental analysis of the PPO-DQN algorithm is conducted in various scenarios, including dynamic obstacles, multi-task switching, and complex terrains. The research results indicate that this algorithm outperforms other classic control algorithms and unoptimized deep reinforcement learning algorithms across various metrics. In terms of path planning, the average path length of PPO-DQN is 21.2 m, significantly shorter than that of PID and A. The average execution time of PPO-DQN is 34.2 s, and the energy consumption is 224.5 J, both significantly better than the control algorithms used for comparison. Additionally, PPO-DQN exhibits outstanding performance in obstacle avoidance success rates, achieving 92.7%, far exceeding the 78.2% of PID and the 83.5% of A. These results demonstrate that the PPO-DQN algorithm can effectively enhance the navigation and control capabilities of robots in complex dynamic environments. In summary, the PPO-DQN algorithm performs exceptionally well in complex environments, providing a more efficient and robust solution for the autonomous control of mobile robots.