Current trends in intra-logistics and industrial manufacturing demand for fleets of mobile robots, or automated guided vehicles (AGVs), respectively, with high degrees of navigation autonomy to realize flexible and adaptive material transportation systems. Thereby, instead of operating in separated cells, the mobile robots are supposed to navigate autonomously in open workspaces, perform reactive obstacle avoidance and interact with pedestrian workers, human-controlled vehicles and other mobile robots. In the context of digitization, industrial environments are becoming more and more networked, raising the question how fleets of mobile robots in these environments can benefit from networked architectures to enhance their navigation capabilities regarding cooperative solutions. For that matter, this work particularly focuses on the aspect of online motion planning, a key component for autonomous mobile robot navigation. The main objective of this work is to develop and implement an online motion planning approach that enables cooperative trajectory optimization when multiple mobile robots are operating in close proximity to each other. In path-crossing scenarios, using a networked navigation architecture, the robots’ local motion planners should be able to cooperatively resolve predicted collisions and obtain smooth transitions with respect to path and velocity profiles for each involved mobile robot. In contrast, current AGV systems usually handle such path-crossing scenarios by implementing static right of way regulations where one or several mobile robots need to stop to let another robot pass. First of all, this thesis proposes a general concept of a networked system architecture for cooperative multi-robot navigation in industrial applications. Based on this concept, a concrete motion planning architecture is derived that enables cooperative optimization in multi-robot path-crossing scenarios by sharing predicted trajectories at runtime between the involved mobile robots. For this purpose, this work introduces a novel online motion planning approach for mobile robots to perform predictive trajectory planning in dynamic environments. In terms of reactive obstacle avoidance, the approach is able to incorporate predicted trajectories of dynamic obstacles. Vice versa, the planning approach considers the individual kinematic and kinodynamic constraints of the mobile robot in order to reliably predict its movement over time. Using the networked architecture, trajectories predicting the robot’s current motion are shared at runtime with other mobile robots to enable cooperative collision avoidance in path-crossing scenarios. The concepts and algorithms presented in this work have been successfully implemented in corresponding software components and have been demonstrated using a set of industrial mobile service robots. Based on this implementation, practical feasibility and performance of the proposed approach are validated and evaluated in a series of single- and multi-robot experiments.


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Predictive and Cooperative Online Motion Planning: A Contribution to Networked Mobile Robot Navigation in Industrial Applications



    Erscheinungsdatum :

    2019-01-01


    Anmerkungen:

    Fraunhofer IPA



    Medientyp :

    Hochschulschrift


    Format :

    Elektronische Ressource


    Sprache :

    Englisch



    Klassifikation :

    DDC:    629