Design, modeling and control of modular tendon-driven elastic continuum mechanisms
The aim of the project is to gain fundamental knowledge about the design, modeling and control of modular tendon-driven elastic continuum mechanisms (ECM's). A modular ECM is composed of multiple modules and is actuated by tendons guided along these modules. To assemble modules to an entire mechanism, e.g. by stacking modules on top of each other, interfaces between the modules will be made available. The interfaces between soft and hard material within the modules will be designed to resist the expected strains. The design should allow for rapid manufacturing and assembly processes. A single module will enable the fundamental motions like pitch, yaw and roll. These motions shall be induced within the assembled ECM by guided tendons being actuated, coupled with each other or fully passive. We want to understand the effect of tendon arrangements and actuation synergies to design suitable tendon routings for applications where high dexterity of the ECM is required. One of the main goals of the project is to use a unified modeling procedure to obtain models for workspace design and models for model-based control. For the workspace design of an ECM, we work out static and dynamic ECM-models. They consist of rigid bodies, nonlinear beams and external configuration dependent forces, which model the effects of the tendons and various other appearing forces. The beam formulation requires, possibly nonlinear, constitutive laws for which an experimental setup will be realized. For the numerical computation, beam finite elements are developed and their performance with respect to precision and computational efficiency will be addressed. The morphology of the system provides an inherent stability with respect to its undeformed configuration. This enables control approaches which can satisfy alternative tasks, as for instance increasing the overall performance of the system or adapting to additional constraints. For the strongly underactuated system highly dynamic motions will be considered. To reach this aim we suggest an approach combining a model-based feedforward controller with a nonlinear feedback. A challenge that arise in this context is, on the one hand, to merge these two control strategies and, on the other hand, to observe the high dimensional state of the mechanical system required for such a control approach.