AUTONOMOUS LANDING AND CONTROL OF QUADRIRROTORS IN GPS-NULL ENVIRONMENTS USING MACHINE LEARNING
Navigation and control of quadrirotors in denied GPS environments is an issue of important discussion, since the GPS system is inherent to most navigation and control methods of quadrirotors, making it very difficult to perform missions in GPS-null environments with conventional methods navigation. This work will discuss a speed control system and an autonomous landing system, both based on machine learning. These systems enable a quadrotor to navigate and land on a specific target, without information about its current position or the target's position, using only sensor data and an on-board camera. The results obtained are promising, where a robust control was obtained, comparable to classic controllers, such as proportional derivative controls and linear quadratic regulators, while the autonomous landing system is capable of converting 93% of the random initial state episodes into successful landings.