Planning of Workplaces with Multiple Kinematically Redundant Robots

Abstract

This thesis provides new methods for planning and optimization of robotic workplaces,
\ni.e. workplaces with assistance of a robotic system. The robots may work autonomously
\nor in cooperation with man. The thesis covers cooperation between multiple robots and
\nlays stress on robots with kinematic redundancy. As highly demanding application area,
\noptimal design and preoperative planning for minimally invasive and open robotic surgery
\nis chosen.
\nOptimization and planning of robotic workplaces are important instruments to cope with
\nthe increased complexity of today’s robotic applications and to ensure safe operation.
\nAlgorithms and devices to facilitate and partially automatize planning and optimization
\nare, however, barely existent so far – with currently available tools the user mostly has to
\nresort to a trial and error approach. Especially for complicated tasks requiring e.g. several
\nrobots cooperating in an unstructured environment, it is very improbable that a good (if
\nat all sufficient) setup of the robotic workplace can be found this way. Therefore, the
\ninclusion of algorithms to automatize the planning procedure as presented in this thesis
\nis the evident next step to be taken.
\nClosed form solutions for inverse kinematics and singularities provide the core of a reliable
\nworkplace optimization system and are developed in this thesis for serial kinematically
\nredundant robots. Unlike state of the art methods, the presented inverse kinematics computation
\ndoes not suffer from algorithmic singularities. The thesis describes an accordingly
\ndeveloped software library for inverse kinematics and shows both a planning procedure
\ninvolving the medical robot KineMedic, and a real-time application, providing inverse
\nkinematics for Cartesian control of the robotic system Justin with a computation time of
\nlower than 0.6 ms for all 18 considered joints.
\nBased on the closed form solutions, the thesis presents a complete procedure for workplace
\noptimization and robot synthesis that uses a two-step algorithm based on Genetic Algorithms
\nand a subsequent Sequential Quadratic Programming method. The thesis develops
\nseveral optimization criteria and demonstrates the performance of the methods with a
\npreoperative planning procedure as well as with the kinematic design optimization of the
\nKineMedic system.
\nThe algorithms implemented in this thesis help the human during the decision taking
\nprocedure, by e.g. providing a preselection of (according to the chosen criteria) good
\nsolutions or by carrying out an optimization in a certain subspace while leaving the determination
\nof the remaining parameters to the human. The thesis facilitates the transfer
\nof a chosen setup into the real environment using a handheld contact-free surface-based
\nregistration procedure with an overall worst case error of 3 mm and a new handheld device
\nto automatically project relevant structures into the real environment.