In recent years, deep reinforcement learning (DRL) approaches have generated highly successful controllers for a myriad of complex domains. However, the opaque nature of these models limits their applicability in aerospace systems and sasfety-critical domains, in which a single mistake can have dire consequences. In this paper, we present novel advancements in both the training and verification of DRL controllers, which can help ensure their safe behavior. We showcase a design-for-verification approach utilizing k-induction and demonstrate its use in verifying liveness properties. In addition, we also give a brief overview of neural Lyapunov Barrier certificates and summarize their capabilities on a case study. Finally, we describe several other novel reachability-based approaches which, despite failing to provide guarantees of interest, could be effective for verification of other DRL systems, and could be of further interest to the community.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Safe and Reliable Training of Learning-Based Aerospace Controllers


    Beteiligte:
    Mandal, Udayan (Autor:in) / Amir, Guy (Autor:in) / Wu, Haoze (Autor:in) / Daukantas, Ieva (Autor:in) / Newell, Fletcher Lee (Autor:in) / Ravaioli, Umberto (Autor:in) / Meng, Baoluo (Autor:in) / Durling, Michael (Autor:in) / Hobbs, Kerianne (Autor:in) / Ganai, Milan (Autor:in)


    Erscheinungsdatum :

    29.09.2024


    Format / Umfang :

    1188106 byte





    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch



    Reliable rubber for aerospace

    English, L.K. | Tema Archiv | 1989


    Fuzzy Logic Controllers for Aerospace Applications

    Sreenatha, A. G. / International Association of Science and Technology for Development | British Library Conference Proceedings | 2001


    SHERPA: a safe exploration algorithm for Reinforcement Learning controllers

    Mannucci, Tommaso / Van Kampen, Erik-Jan / de Visser, Coen C. et al. | AIAA | 2015


    Hierarchically Structured Controllers for Safe UAV Reinforcement Learning Applications

    Mannucci, Tommaso / Van Kampen, Erik-Jan / de Visser, Coen C. et al. | AIAA | 2017


    Safe Reinforcement Learning Benchmark Environments for Aerospace Control Systems

    Ravaioli, Umberto J. / Cunningham, James / McCarroll, John et al. | IEEE | 2022