MMLT/ik: Efficiently Learning Mealy Machines with Local Timers by Using Imprecise Symbol Filters

Active automata learning (AAL) can infer accurate automata models of real-time systems (RTS). However, even efficient AAL methods for RTS, like learning Mealy machines with local timers (MMLTs), take very long to infer models with large input alphabets. We introduce MMLT/ik, a method for learning accurate MMLTs fast by ignoring transitions based on imprecise prior knowledge (ik) from an imprecise symbol filter. We validate our method across a diverse set of case studies with automotive system components, network protocols, wireless sensor networks, and smart home appliances, where we reduce runtime drastically.