Enhancing Autonomous Navigation Through Hybrid DQN Optimization: Integrating Prioritized Replay and Simulated Annealing in Dynamic Environments

Abstract Autonomous navigation in unstructured environments remains a critical challenge for mobile robotics, where conventional planners like A* and potential field methods suffer from limited adaptability in dynamic scenarios. This study develops an enhanced Deep Q-Network (DQN) architecture incorporating three synergistic innovations: 1) A prioritized experience replay mechanism employing TD-error weighted sampling to accelerate policy convergence (38.7\% faster than vanilla DQN); 2) A hybrid action space framework combining discrete directional choices (8 headings) with continuous velocity control (0-2 m/s resolution), enabling collision avoidance under 0.3m proximity constraints; 3) A simulated annealing-inspired exploration strategy with adaptive temperature decay , dynamically balancing exploration-exploitation tradeoffs. Extensive experiments across six benchmark environments demonstrate the proposed method's superiority, achieving 21.3\% shorter path lengths than RRT* and 63.4% higher success rates than traditional DQN in cluttered dynamic settings. Particularly noteworthy is its 92.7% computational efficiency improvement over A* in real-time replanning tasks. These advancements establish a robust foundation for deploying autonomous robots in logistics and disaster response applications.