Recent Papers

Autonomous drone navigation in complex urban environments presents significant challenges due to dynamic obstacles, GPS signal occlusions, and high-density layouts. Deep Reinforcement Learning (DRL) offers a promising approach for enabling drones to learn navigation policies that adaptively respond to complex real-world conditions without explicit programming. This research investigates the application of DRL algorithms, specifically Deep Q-Networks (DQN) and Proximal Policy Optimization (PPO), to enable autonomous drone path planning and collision avoidance in simulated urban environments. The study evaluates training efficiency, policy robustness, and real-time performance, comparing DRL methods with classical rule-based navigation. Results indicate that DRL agents achieve higher success rates in navigating urban mazes with dynamic obstacles, demonstrating adaptability and potential for real-world deployment. Challenges such as reward shaping, sample efficiency, and sim-to-real transfer are discussed, with recommendations to enhance training strategies and scalability.Keywords: Autonomous drone navigation, Deep Reinforcement Learning (DRL), Deep Q-Network (DQN), Proximal Policy Optimization (PPO), path planning, collision avoidance, and sim-to-real transfer.

Copyright © 2025 Geethanjali N R. This is an open access article distributed under the Creative Commons Attribution License.