Conditional Imitation learning is a common and effective approach to train autonomous driving agents. However, two issues limit the full potential of this approach: (i) the inertia problem, a special case of causal confusion where the agent mistakenly correlates low speed with no acceleration, and (ii) low correlation between offline and online performance due to the accumulation of small errors that brings the agent in a previously unseen state. Both issues are critical for state-aware models, yet informing the driving agent of its internal state as well as the state of the environment is of crucial importance. In this paper we propose a multi-task learning agent based on a multi-stage vision transformer with state token propagation. We feed the state of the vehicle along with the representation of the environment as a special token of the transformer and propagate it throughout the network. This allows us to tackle the aforementioned issues from different angles: guiding the driving policy with learned stop/go information, performing data augmentation directly on the state of the vehicle and visually explaining the model's decisions. We report a drastic decrease in inertia and a high correlation between offline and online metrics.


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Addressing Limitations of State-Aware Imitation Learning for Autonomous Driving



    Erscheinungsdatum :

    2024-01-01



    Medientyp :

    Aufsatz (Zeitschrift)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch



    Klassifikation :

    DDC:    629



    Addressing Limitations of State-Aware Imitation Learning for Autonomous Driving

    Cultrera, Luca / Becattini, Federico / Seidenari, Lorenzo et al. | IEEE | 2024

    Freier Zugriff

    Dynamic Conditional Imitation Learning for Autonomous Driving

    Eraqi, Hesham M. / Moustafa, Mohamed N. / Honer, Jens | IEEE | 2022


    Hierarchical Interpretable Imitation Learning for End-to-End Autonomous Driving

    Teng, Siyu / Chen, Long / Ai, Yunfeng et al. | IEEE | 2023


    Imitation Learning for Autonomous Vehicle Driving: How Does the RepresentationMatter?

    Greco, Antonio / Rundo, Leonardo / Saggese, Alessia et al. | British Library Conference Proceedings | 2022


    Energy-Aware Driving Profile of Autonomous Electric Vehicles Considering Regenerative Braking Limitations

    Mohammadi, Masoud / Heydari, Shoeib / Fajri, Poria et al. | IEEE | 2022