Low-order Linear Estimation of a Commercial Quadrotor using Flight Test Data Sets

In this paper, a functional model, which is identified based on flight test data sets, is used to improve the performance of guidance loop of a commercial quadrotor equipped with a low-cost autopilot and navigation system. For this purpose, relationships between horizontal velocities and the quadrotor attitudes around the horizontal axes and relationships between three quadrotor attitudes with the correspondent commands sent to the autopilot are identified independently. The first model describes the relation between rotation around the axes of the quadrotor with declared commands in the autopilot system and the other describes the relation between horizontal velocities with the angle around the longitudinal and latitudinal axes of the quadrotor. According to the experimental test results, a first order transfer function has enough accuracy to describe the rotation around the vertical axis with related commands. Also, the description of the relation between the horizontal velocities and the autopilot commands has suitable accuracy using two consequent first order transfer functions; one model between the applied commands and changes of roll and pitch angles, and the other between roll and pitch angles and the horizontal velocities. Due to the linear time invariant characteristics of the system, the least square estimator is used to estimate unknown parameters of the models. To evaluate the identified models, the performance of the path tracking using a PID controller as the outer controller is investigated. The results of the indoor tests show significant improvements in path tracking that confirms enough accuracy of the identified models.