This thesis studies the feasibility of outdoor agile autonomous flight of a quadrotor that is equipped with an RTK-GPS, IMU and magnetometer.
The novelty of this study is the autonomous flight of agile trajectories using low sample rate pose measurements. The open-source autonomous agile flight platform Agilicious is used as starting point, where the bridges and state estimator are altered. An extended Kalman filter is used as a state estimation algorithm, and model predictive control is used as the control algorithm. GPS samples at a low sample rate of 5 Hz are fused with IMU samples at a high sample rate of 400 Hz in the extended Kalman filter. The feasibility is tested using a closed-loop simulation, where sensor data is generated using noise characteristics based on the sensors.
The results are compared to a benchmark that uses perfect state estimation. From the results, it is found that the implementation has the same performance as the benchmark flying trajectories with accelerations of over 3g, angular rates of 700 deg/s and velocities of over 60 km/h. Furthermore, there is leeway in the maximum noise levels of the sensor messages. These results indicate a strong possibility that this implementation will work in real-life flights.