Navigation in Closed Environments of a Quadrirrotor Vehicle Using Computer Vision
The autonomous navigation of four rotors in closed environments has become an area of interest, because in these environments the Global Positioning System (GPS) has its functioning compromised, sometimes making its use impossible. Since navigating the vehicle requires knowledge of its location, the absence of a GPS makes it impossible for the vehicle to fly autonomously. This work proposes a system for estimating the position and attitude of a four-rotor vehicle using computer vision, which will be used to enable vehicle navigation in closed environments without GPS. The dynamic model of the quadrirotor is obtained through the Newton-Euler formalism, which provides the basis for the design of a Proportional Derivative (PD) control that enables vehicle navigation through a planned trajectory using the snap minimization approach. Through the use of digital image processing techniques, the vision system is able to determine the vehicle's attitude and position. In addition, the information obtained by visual sensing is combined with inertial sensors through the Extended Multiplicative Kalman Filter (FKME), in order to obtain more accurate estimates about the vehicle's orientation. The system is validated through a 3D simulation environment, which has cameras fixed outside the vehicle, artificial markers that help determine the vehicle's states, in addition to the three-dimensional model of the four-rotor vehicle. The results obtained showed that the proposed position and attitude estimation system is able to provide the states with satisfactory errors for the desired application. With the estimated states, the quadrirotor was able to successfully track the desired trajectories, validating the proposal of the developed system.