Modelling, Control and Optimization of Modular Reconfigurable Robots

Modular reconfigurable robots are robotic systems offering new opportunities to rapidly create fit-to-task flexible automation lines. The recent trends of increasingly varying market needs in low-volume high-mix manufacturing demands for highly adaptable robotic systems like this. In this context, methods for quickly and automatically generating a modular robot model and controller should be developed. Moreover, modularity and reconfigurabilty open up new opportunities for on-demand robot morphology optimization for varying tasks. Therefore a method to optimize the robot design for a certain criterion should be provided in order to exploit the full potential of reconfigurable robots. In this thesis, a complete hard- and software architecture for a modular reconfigurable EtherCAT-based robot is presented. This novel approach allows to automatically reconstruct the topology of different robot structures, composed of a set of body modules, each of which represents an EtherCAT slave. This approach enables to obtain in an automatic way the kinematic and dynamic model of the robot and store it in URDF format as soon as the physical robot is assembled or reconfigured. The method also automatically reshapes a generic optimization-based controller to be instantly used after reconfiguration. Finally, a study and analysis on how to find the best suited reconfigurable robot morphology for a given task are presented, starting from a fixed set of joint and link modules. In particular, is shown how exploiting multi-arm robotic systems and modifying the relative and absolute positions of their bases, can expand the solution space for a given task. Results obtained in simulations for different tasks, are verified with real-world experiments using a in-house developed reconfigurable robot prototype.