Machine learning-based techniques for the prediction of complete coordination of astrobots swarms.

Astrobots are robotic manipulators present in the latest generation telescopes. Their role is to coordinate optical fibers in specific positions of a focal plane, so that it is possible to get astronomical information of a desired observation. The crowded arrangement of these manipulators inside the focal plane of a telescopes makes the coordination process so much complicated that it is not always possible for a single astrobot to reach its target configuration. Therefore the convergence verification before the execution of a coordination is of particular interest. However a formal method for this purpose has not yet been found in the past. The main objective of this thesis is to develop formal methods for the prediction of convergence of astrobots swarm using some machine learning techniques. In particular two prediction algorithms have been developed, a KNN-based algorithm and an SVM-based algorithm. Furthermore, a possible approach to solving the problem is proposed which exploits the use of a convolutional neural network.