UAV Obstacle Avoidance with Moving Object Prediction for Safe Flight Using Deep Reinforcement Learning

With the development of technology, the use of reinforcement learning for autonomous flight of Unmanned Aerial Vehicles (UAVs) is becoming more and more widespread. The detection and avoidance of obstacles will be an important issue when UAVs are performing missions in unknown environments. In order to achieve safe autonomous flight, we want to predict the direction vector of the moving objects by particle filtering and find the safety value between any point on the plane and the objects to determine the safety of the UAV flight. Next, we determine the distance between the safety value and the target point, such as whether the safety value of the selected location is sufficient, the distance between the UAV and the moving objects, and the distance between the UAV and the target point. These conditions are used to design our reinforcement learning model. Using the training results of the reinforcement learning, we give the information about flight safety to the UAV for use. The UAV can choose a safe position through autonomous flight and keep the UAV flying to the destination in a safe and collision-free condition.