From Passive to Active: Learning Timed Automata Efficiently

Bernhard Aichernig, Andrea Pferscher*, Martin Tappler

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference paperpeer-review


Model-based testing is a promising technique for quality assurance. In practice, however, a model is not always present. Hence, model learning techniques attain increasing interest. Still, many learning approaches can only learn relatively simple types of models and advanced properties like time are ignored in many cases. In this paper we present an active model learning technique for timed automata. For this, we build upon an existing passive learning technique for real-timed systems. Our goal is to efficiently learn a timed system while simultaneously minimizing the set of training data. For evaluation we compared our active to the passive learning technique based on 43 timed systems with up to 20 locations and multiple clock variables. The results of 18060 experiments show that we require only 100 timed traces to adequately learn a timed system. The new approach is up to 755 times faster.
Original languageEnglish
Title of host publicationNASA Formal Methods - 12th International Symposium, NFM 2020, Proceedings
Subtitle of host publication 12th International Symposium, NFM 2020, Moffett Field, CA, USA, May 11-15, 2020, Proceedings
EditorsRitchie Lee, Susmit Jha, Anastasia Mavridou
Number of pages19
ISBN (Print)978-3-030-55753-9
Publication statusPublished - 10 Aug 2020
Event12th NASA Formal Methods Symposium: NFM 2020 - NASA Ames Research Center, Moffett Field, United States
Duration: 12 May 202014 May 2020

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12229 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349


Conference12th NASA Formal Methods Symposium
Abbreviated titleNFM 2020
Country/TerritoryUnited States
CityMoffett Field
Internet address


  • Active automata learning
  • Genetic programming
  • Timed automata
  • Model learning
  • Model inference

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fields of Expertise

  • Information, Communication & Computing


Dive into the research topics of 'From Passive to Active: Learning Timed Automata Efficiently'. Together they form a unique fingerprint.

Cite this