Automatic generation of task-specific serial mechanisms using combined structural and dimensional synthesis

摘要

Although the diversity of robot kinematic structures that are used in industry has increased in the recent years, robot manufacturers still offer only a limited number of architectures. As most of the robots, especially in industrial applications, are chosen from a few different kinematic structures, they are oversized as well as overactuated and thus not optimal with respect to their task. To develop new robot manipulators, two processes have to be considered, namely the structural and the dimensional synthesis. The former allows for determining the kinematic structures that fulfill a given number of degrees of freedom at the end effector. In the second, the dimensions of the robot links are optimized with respect to a desired performance index. However, these processes are actually performed separately. In the dimensional synthesis, only one structure is usually optimized without consideration of possible further mechanisms that are able to perform the desired task. Consequently, the obtained manipulator is not necessarily optimal with respect to its application. This thesis presents a new approach for the synthesis of task specific serial manipulators considering all suitable structures. Initially, the task suitable structures as well as their optimization parameters are automatically determined from all Denavit parameters combinations by inspecting the rank of the Jacobian matrix and the end effector velocity in symbolic form. Afterwards, isomorphisms are detected through the dependency of the desired degrees of freedom with respect to the geometric parameters. As a result, the number of structures and optimization parameters is significantly reduced allowing for the optimization of all task suitable structures. In order to be able to calculate kinematic and dynamic performance indices, the generated robotic structures are modeled and executable code is automatically generated. This code is individual for each architecture and offers a reduced computational effort compared with numeric methods. The performances indices are used to constitute the objective function and the constraints of the optimization problem. The task specific manipulator results from the geometric optimization of all suitable structures. The approach can synthesize optimal manipulators up to six degrees of freedom. As most of the method is automatically performed, it is a powerful tool for reducing the effort during the conceptual design phase of a robot. Additionally, it allows for the comparison of the performance between the structures as well as for their sensitivity with respect to changes in the optimal geometric parameters.