Active Learning of Modular Plant Models

Automatically Learning Formal Models from Autonomous Driving Software

“…The Modular Plant Learner [30] is a state-based active learning algorithm developed to learn a modular model, that is, one composed of a set of interacting automata. These modules together define the behavior of the system.…”

“…It exploits structural knowledge of the SUL to search smartly. Hence, it requires access to the SUL's variables, and a Plant Structure Hypothesis (PSH) [30] defining the structure of the system. The PSH is a 3-tuple H = D, E, S , where D is a set that provides a unique name for each module that is to be learned.…”

“…An automaton representing their joint behavior is given in Figure 3c. The L * algorithm [26] is a well-known active automata learning algorithm th 239 has inspired a tremendous amount of work [27][28][29][30]33,36,46,52]. It learns a minim 240 automaton M that generates the regular language L( M) ⊆ Σ * representing the behav 241 of the SUL.…”

“…This depends on the user-defined PSH and the possibility to decouple the system. Further details about the MPL are found in [30]. bmitted to Electronics 10 of 22 ⟨i,i⟩ ⟨w,i⟩ ⟨i,w⟩ ⟨w,w⟩…”