Model-Based Control for Industrial Robots: Uniform Approaches for Serial and Parallel Structures

摘要

Nowadays, there are still two major challenges for industrial robotics in automated production. These are enhancing manufacturing precision and reducing cycle-times. Beside the advances made in the classic robotics over the last decades, new technologies are emerging in the industrial field aiming more flexible high-speed and accurate manufacturing. Robots with parallel structures are attracting the attention of automati on industry as an innovative product with high dynamic potentials. Such robots, like the tricpet are integrated nowadays by BMW, Volvo or Airbus in their manufacturing lines. Compared to each other, the classic serial (or open) chain robots and the parallel (or closed) chain robots have their specific benefits and suffer from own drawbacks. The proposed chapter gives a comparison of the two types in the scope of their suitability for solving modern problems in industrial robotics. Additionally, appropriate approaches are proposed to remove drawbacks of classic industrial control solutions. Hereby, it is focussed on model-based strategies for ameliorating control accuracy at high dynamics and therefore to expose solutions towards high-speed automation. One of the main purposes of the proposed chapter is to contribute to extending the state of the art in industrial robotics by the innovative class of parallel robots. Furthermore, classic and advanced model-based control approaches are discussed for both robot types. Uniform methodologies for both classes are given. It is focused on crucial issues for practical application in the industrial filed. The first aspect is surely the modelling of kinematics (see section 2) and dynamics (see section 3) for serial and parallel robots. Here, an opposite duality in formalism is shown. By appropriate choice of minimal coordinates and velocities, the inverse dynamics of the two robot classes can be derived by the principle of virtual power. This yields computational highly efficient models that are well appropriate for real-time applications. Since the success of such feedforward control depends on the estimation quality of the model parame