A Novel Nonlinear Model Predictive Control for coordinating a UAV Swarm

This thesis project addresses the problem of coordinating the motion of a multiagents system for a real time application. The case study is a Unmanned Aerial Vehicles (UAV) swarm moving on a not completely known map with pop-up obstacles to be avoided. The main concept of coordination is discussed, providing a general procedure to describe the environment and mission goals through a proper optimization problem. Relative distances are used to describe the tasks, as in a standard formation control. The usage of a minimal task description permits the swarm to have more degree of freedom to achieve the optimal configuration. Dealing with not completely known environments, where the robot has to react to sensory data as quickly as possible, a real time algorithm is necessary. For this reason, the controller design needs to properly face the computational burden and environments variations issues. Furthermore, the algorithm should be distributed. The report presents a novel approach that allows to easily deal with nonlinear models and very complex, time variant cost function. A distributed Non Linear Model Predictive Control is implemented and a Particle Swarm Optimization is used to solve the optimization problem