Trajectory Generation for a Quadrotor Unmanned Aerial Vehicle
Author | : Douglas Conover |
Publisher | : |
Total Pages | : 96 |
Release | : 2018 |
ISBN-10 | : OCLC:1121604221 |
ISBN-13 | : |
Rating | : 4/5 (21 Downloads) |
Download or read book Trajectory Generation for a Quadrotor Unmanned Aerial Vehicle written by Douglas Conover and published by . This book was released on 2018 with total page 96 pages. Available in PDF, EPUB and Kindle. Book excerpt: The field of multirotor unmanned aerial vehicles (UAVs) has seen substantial progression in the past decade. Trajectory generation and control has been a main focus in this domain, with methods that enable the performance of complex three-dimensional maneuvers through space. Efforts have been made to execute these maneuvers using concepts of nonlinear control and differential flatness. However, a lack of theory for the estimation of higher-order dérivatives of a multirotor UAV has prevented the experimental application of several of these techniques concentrated on trajectory control. This work firstly explores the existing control approach of sequential composition for the execution of quadrotor manoeuvres through narrow windows. This technique involves the combination of several theoretically simple controllers in sequence in order to produce a complex result. Experimental results conducted in the Mobile Robotics and Automated Systems Laboratory (MRASL) at Polytechnique demonstrate the validity of this approach, producing precise and repeatable manoeuvres through narrow windows. However, they also show the limitations of such a method in real world applications, notably its initial inaccuracy and lack of feasibility evaluation. This thesis then focuses on the development of a state-estimation architecture based on linear Kalman filter techniques in order to provide a real-time value of a quadrotor UAV's second and third derivatives (referred to as acceleration and jerk, respectively). Filters of different complexities are developed with the goal of incorporating all available system information into the resulting estimate. A full-state estimator is produced that uses a quadrotor's position and acceleration measurements as well as control inputs in order to be usable for feedback. A jerk-augmented controller based off of optimal control theory is then developed in order to validate this estimator. It is designed in such a way to use the UAV's jerk, acceleration, velocity and position as design parameters and to be unstable without feedback in each of these terms. Tests are conducted in order to examine the performance of both the estimator and controller. Firstly, the quadrotor is commanded to track various reference inputs in 3D space to ensure its stability. The controller tracks these references very closely to simulated responses. The controller is then asked to follow a changing reference in order to evaluate the precision of the developed estimator. Results show that the real-time estimation of the jerk follows offline values adequately. To the best of our knowledge, this is the first application to implement the feedback of a multirotor UAV's jerk in real-world experimentation.