Autonomous Drone Navigation Using Reinforcement Learning and Real-Time Sensor Data Fusion

Abstract

The increasing adoption of drones across domains such as surveillance, logistics, agriculture, and disaster response has intensified the need for reliable autonomous navigation systems. This work presents an adaptive navigation framework that integrates reinforcement learning (RL) with real-time multi-sensor data fusion. Unlike conventional rule-based approaches, the proposed method enables the unmanned aerial vehicle (UAV) to learn from continuous interaction with its environment. Data from GPS, IMU, LiDAR, and vision sensors are combined to construct a dynamic representation of the surroundings. Based on this representation, the RL agent incrementally improves its navigation strategy by evaluating past actions and outcomes. Experimental observations indicate improved adaptability, enhanced navigation accuracy, and more effective obstacle avoidance compared to traditional techniques.