Testing and Active Learning of Resettable Finite-State Machines

This thesis proposes novel active-learning algorithms and testing methods for deterministic finite-state machines that (i) have a specified transition from every state on each input of the (fixed) alphabet and (ii) can be reliably reset to the initial state on request. These algorithms rely on the novel methods of construction of separating sequences. Extensive evaluation demonstrates that the described testing and learning methods are the most efficient in terms of the amount of interaction by a tester with the system under test.