Having robots operating in unstructured and dynamically changing environments is a challenging task that requires advanced motion generation approaches that are able to perform in real-time while maintaining the robot and environment safety. The progress in the field of numerical optimization, as well as the development of tailored algorithms, made Nonlinear Model Predictive Control (NMPC) an appealing candidate for real-time motion generation. By considering the robot model as prediction model and through appropriate constraints on the robot states and control inputs, NMPC can enforce safety to the resulting motion in a straightforward way. This thesis addresses the problem of real-time generation of safe motions for mobile robots and mobile manipulators. The different structure of the considered robots introduces different safety risks during the robot motion and so the motion generation problem for each robot is addressed in separate parts of this thesis. In the first part, the problem of motion generation for mobile robots navigating in environments populated by static and/or moving obstacles is considered. For the generation of the desired motion, real-time NMPC is used. We argue that, in order to tackle the risk of collision with the environment, traditional distance-based approaches are incapable of maintaining the robot safety when the NMPC uses relatively short prediction horizons. Instead, we propose two NMPC approaches that employ two alternative collision avoidance constraints. The first proposed NMPC approach is applied to a scenario of safe robot navigation in a human crowd. The NMPC serves as a motion generation module in a safe motion generation framework, complete with a crowd prediction module. The considered collision avoidance constraint is built upon an appropriate Control Barrier Function (CBF). The second NMPC approach is applied to a scenario of robot navigation among moving obstacles, where the dynamics of the considered robot are significant. The proposed collision avoidance constraint ...


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Optimization-based methods for real-time generation of safe motions in mobile robots



    Erscheinungsdatum :

    2023-01-25


    Medientyp :

    Hochschulschrift


    Format :

    Elektronische Ressource


    Sprache :

    Englisch



    Klassifikation :

    DDC:    629



    Moving Personal Robots in Real-Time Using Primitive Motions

    Xu, L. / Zheng, Y. F. | British Library Online Contents | 2001


    Safe Path Planning for Mobile Robots

    Lambert, Alain / Piat, Nadine Le Fort | ASCE | 1998




    cRRT*: Planning loosely-coupled motions for multiple mobile robots

    Rosell Gratacòs, Jan / Suárez Feijóo, Raúl | BASE | 2014

    Freier Zugriff