ARCHITECTURE OF AUTONOMOUS CONTROL SYSTEM FOR FPV-DRONES

Abstract

Drone technologies and robotic systems, that can be used to optimize or replace human labor in different areas, are continuing to develop and expand their area of application. Computational hardware, that can be used as an extension to basic drone hardware, is becoming more powerful for a smaller price and size. Autonomous drone control provides a practical solution for issues that are present when a drone is controlled manually by an operator. There are limitations that exclude the possibility of operator control and some solutions are much less effective if there is no autonomous algorithm. A large distance, radio interference or a huge amount of data to be analyzed are not as critical if a drone is capable to make decisions autonomously without an operator. Because of the features of a basic amateur FPV-drone its flight controller does not operate enough computing power to execute complex algorithms such as object detection, object tracking, calculation of a proper trajectory etc. In cases like this a companion computer should be used. Depending on an area of application, drone control algorithms may vary. Flight controllers’ firmware has some embedded functionality with relatively simple algorithms like pre-defined flight routes that use GCS-coordinates. In the case when more complex behavior is required (image processing, AI integration or autonomous decision making with manual controls), such firmware provides an ability to accept various drone commands that correspond to the MAVLink communication protocol. This work investigates important details of companion computers integration with flight controllers with examples and describes hardware and software aspects of implementation of autonomous drone controls. The research also provides solutions to problems that can be possible when implementing such a system.