Design of a small quadrotor UAV and modeling of an MPC-based simulator

The concept of Autonomous Vehicles (AVs), such as self-driving cars and Unmanned Aerial Vehicles (UAVs), has been widely explored in recent years. The aims of this thesis are to develop, implement and test, both hardware (HW) and software (SW) system of a multirotor UAV for indoor applications. There are many quadrotors that are commercially available nowadays, however, off-the-shelf quadrotors usually lack the ability to be reprogrammed and are unsuitable for use as research platforms. Therefore, an HW selection has been carried out in the first part of the project, starting from Commercial-Off-the-Shelf (COTS) components.. The quadrotor autopilot is based on ‘Pixhawk 4 mini’ and it is able to fly indoor using stationary ultrasonic beacons (‘Marvelmind Set HW v4.9’) as GPS. Using personalized 3D printed components, all the selected items were integrated and assembled in an efficient, working configuration. The first part of the project ended with the verification and validation of communication between HW’s components and flying tests. The second part of the project concerns the SW development and testing. This thesis proposed a Receding Horizon Control (RHC). RHC, also known as Model Predictive Control (MPC), is an optimal control-based strategy that uses a plant model to predict the effect of an input profile on the evolving state of the plant. At each sample time, an optimal control problem is solved, and its optimal input profile is implemented until another plant measurement becomes available. The updated information is used to formulate and solve a new optimization problem, and the process is repeated. A practical disadvantage is its computational cost, so the MPC applications are usually limited to linear process with relatively slow dynamics such as chemical engineering systems. However, with the raising of more powerful computers it is now being used in system with faster dynamics. Quadrotor dynamics can be split into two categories: slow dynamics, regarding the position; and fast dynamics, regarding the attitude and altitude. An MPC-based cascade controller is developed. The controller is structured into two loops: an outer loop related to UAVs slow dynamics (controlled by a PD controller) and an inner loop related to UAVs fast dynamic (controlled by MPC controller). The controller is tested tracking different waypoints-following trajectories with increasing difficulty in terms of changes in the states. In these studies, a linearized model is adopted, and the Receding Horizon method is applied to generate the optimal control sequence. Several simulations are conducted to examine and evaluate the performance of the proposed control approach using MATLAB and Simulink environment. Simulation results show that this kind of control is highly effective to track different types of given reference trajectory. The performances of the controller have been further tested on a virtual environment, using Unreal Engine as plant to have more realistic results representations. Future work involves Hardware-in-the-Loop (HIL) and Processor-in-the-Loop testing. Once such processes are concluded, a custom MAVLink-based flight controller (C language) will be developed, deploying it into the real quadrotor flight computer and testing the algorithm’s performance with experimental flight tests.