A Hybrid Deep Reinforcement Learning Approach for Autonomous Drone Navigation in Dynamic Environments

Research Article

Abstract

Autonomous navigation of drones in dynamic environments remains a significant challenge due to the need for real-time decision-making, obstacle avoidance, and environmental adaptability. This paper proposes a hybrid deep reinforcement learning (DRL) framework that combines model-based planning with model-free learning to enable robust and efficient drone navigation. The model leverages Convolutional Neural Networks (CNNs) for spatial feature extraction and Long Short-Term Memory (LSTM) networks for temporal decision-making, integrated within a Proximal Policy Optimization (PPO) framework. Simulation results in dynamic urban and forest scenarios demonstrate improved performance in terms of navigation success rate, collision avoidance, and learning efficiency compared to traditional DRL methods.