Uncontrolled intersections are important and challenging traffic scenarios for autonomous vehicles. Vehicles not only need to avoid collisions with dynamic vehicles instantaneously but also predict their behavior then make long-term decisions in reaction. To solve this problem, we propose a cooperative framework composed of a Primary Driver (PD) for motion planning and a Subordinate Driver (SD) for decision-making. SD is essentially the combination of a prediction module and a high-level behavior planner, which develops a prediction-guided strategy tree to determine the optimal action sequence. Especially, under the guidance of the prediction results, the tree branches are evaluated in security metrics, then get trimmed in action and observation space to reduce the dimensional complexity. With the assistance of SD, PD works as a collision checker and a low-level motion planner to generate a safe and smooth trajectory. We use the INTERACTION dataset to validate our method and achieve more than 90% success rate with efficiency improvement in various situations.


    Zugriff

    Zugriff prüfen

    Verfügbarkeit in meiner Bibliothek prüfen

    Bestellung bei Subito €


    Exportieren, teilen und zitieren



    Titel :

    Decision-making and Planning Framework with Prediction-Guided Strategy Tree Search Algorithm for Uncontrolled Intersections


    Beteiligte:
    Zhang, Ting (Autor:in) / Fu, Mengyin (Autor:in) / Song, Wenjie (Autor:in) / Yang, Yi (Autor:in) / Wang, Meiling (Autor:in)


    Erscheinungsdatum :

    2022-10-08


    Format / Umfang :

    3142089 byte




    Medientyp :

    Aufsatz (Konferenz)


    Format :

    Elektronische Ressource


    Sprache :

    Englisch