Automatic control of a multirotor

Objective of this thesis is to describe the design and realisation phases of a multirotor to be used for low risk and cost aerial observation. Starting point of this activity was a wide literature study related to the technological evolution of multirotors design and to the state of the art. Firstly the most common multirotor configurations were defined and, according to a size and performance based evaluation, the most suitable one was chosen. A detailed computer aided design model was drawn as basis for the realisation of two prototypes. The realised multirotors were “X-shaped” octorotors with eight coaxially coupled motors. The mathematical model of the multirotor dynamics was studied. “Proportional Integral Derivative” and “Linear Quadratic” algorithms were chosen as techniques to regulate the attitude dynamics of the multirotor. These methods were tested with a nonlinear model simulation developed in the Matlab Simulink environment. In the meanwhile the Arduino board was selected as the best compromise between costs and performance and the above mentioned algorithms were implemented using this platform thanks to its main characteristic of being completely “open source”. Indeed the multirotor was conceived to be a serviceable tool for the public utility and, at the same time, to be an accessible device for research and studies. The behaviour of the physical multirotor was evaluated with a test bench designed to isolate the rotation about one single body axis at a time. The data of the experimental tests were gathered in real time using a custom Matlab code and several indoor tests allowed the “fine tuning” of the controllers gains. Afterwards a portable “ground station” was conceived and realised in adherence with the real scenarios users needs. Several outdoor experimental flights were executed with successful results and the data gathered during the outdoor tests were used to evaluate some key performance indicators as the endurance and the maximum allowable payload mass. Then the fault tolerance of the control system was evaluated simulating and experimenting the loss of one motor; even in this critical condition the system exhibited an acceptable behaviour. The reached project readiness allowed to meet some potential users as the “Turin Fire Department” and to cooperate with them in a simulated emergency. During this event the multirotor was used to gather and transmit real time aerial images for an improved “situation awareness”. Finally the study was extended to more innovative control techniques like the neural networks based ones. Simulations results demonstrated their effectiveness; nevertheless the inherent complexity and the unreliability outside the training ranges could have a catastrophic impact on the airworthiness. This is a factor that cannot be neglected especially in the applications related to flying platforms. Summarising, this research work was addressed mainly to the operating procedures for implementing automatic control algorithms to real platforms. All the design aspects, from the preliminary multirotor configuration choice to the tests in possible real scenarios, were covered obtaining performances comparable with other commercial of-the-shelf platforms.