An Improved Monocular Visual Navigation Algorithm for Quadrotor in Indoor Environments

This paper proposes an environment perception algorithm based on the principle of vanishing point in perspective projection and a cascaded multi-variable RBF neural network PID adaptive control method to address the drawbacks of the existing indoor navigation algorithms for quadcopters, such as high computing data and insufficient real-time performance. The control method and the navigation strategy based on vanishing point as the target point are utilized to achieve autonomous indoor flight of the quadcopters. The algorithm is developed on the DJI Robomaster TT drone, and the feasibility and robustness of the algorithm are verified through multiple autonomous flight experiments. The results show that the algorithm has good stability and real-time performance and is suitable for some special environments, meeting the requirements of autonomous indoor flight of the aircraft in practical applications.