UAV Path Planning and Obstacle Avoidance Based on Reinforcement Learning in 3D Environments

This study proposes using unmanned aerial vehicles (UAVs) to carry out tasks involving path planning and obstacle avoidance, explore how improve work efficiency ensure the flight safety of drones. One applications under consideration is aquaculture cage detection; net-cages used in sea-farming are usually numerous scattered widely over sea. It necessary save energy consumption so that drones can complete all detections return their base on land. In recent years, application reinforcement learning has become more extensive. this study, proposed method mainly based Q-learning algorithm enable improvements planning, we compare it with a well-known state–action–reward–state–action (SARSA) algorithm. For avoidance control procedure, same for training AirSim virtual environment; parameters changed, results compared.