Despite the rapid growth in the number of robots in the world, the number of service robots is still very low. The major reasons for this include the robots' lack of world knowledge, sensitivity, safety and flexibility. This thesis experimentally addresses the last three of these issues (sensitivity, safety and flexibility) with reference to advanced, industrial level robotic arms provided with integrated torque sensors at each joint.The aims of this work are twofold. The first one, at a more technical level, is the implementation of a real-time software infrastructure, based on Orocos and ROS, for a general, robust, flexible and modular robot control framework with a relatively high level of abstraction. The second aim is to utilize this software framework for Programming by Demonstration with a class of algorithms known as Dynamic Movement Primitives. Using kinesthetic teaching with one or multiple demonstrations, the robot performs simple sequential in-contact tasks (e. g. writing on a notepad a previously demonstrated sequence of characters). The system is not only able to imitate and generalize from demonstrated trajectories, but also from their associated force profiles during the execution of in-contact tasks. The framework is further extended to successfully recover from perturbations during the execution and to cope with dynamic environments. ; Validerat; 20140822 (global_studentproject_submitter)


    Access

    Download


    Export, share and cite



    Title :

    Programming by Demonstration for in-contact Tasks using Dynamic Movement Primitives


    Contributors:

    Publication date :

    2014-01-01


    Remarks:

    Local 41afe14b-ceeb-4de4-b6a1-42228051a968


    Type of media :

    Theses


    Type of material :

    Electronic Resource


    Language :

    English



    Classification :

    DDC:    629




    Lunar Excavator Mission Operations using Dynamic Movement Primitives

    Joseph M. Cloud / Minh Q. Tram / William J. Beksi et al. | NTRS


    Dynamic Movement Primitives: Volumetric Obstacle Avoidance Using Dynamic Potential Functions

    Michele Ginesi / Daniele Meli / Andrea Roberti et al. | BASE | 2021

    Free access