Numerical robot kinematics based on stochastic and molecular simulation methods

I started to research on this topic in the course of my dissertation under Prof. Dr. Franz Vesely, University of Vienna, in April 1994. In May 1996 our article "Numerical robot kinematics based on stochastic and molecular simulation methods" by Th. Kastenmeier and F. Vesely was printed in the journal Robotica (1996) volume 14 part 3, pp 329-337 (Cambridge University Press, 1996).

Abstract:
Multilink robot arms are geometrically similar to chain molecules. We investigate the performance of molecular simulation methods, combined with stochastic methods for optimization, when applied to problems of robotics. An efficient and flexible algorithm for solving the inverse kinematic problem for redundant robots in the presence of obstacles (and other constraints) is suggested. The algorithm combines the "SHAKE" (Ryckaert 1977), to handle the constraints of the robot arm, with the "SIMULATED ANNEALING" (Kirkpatrick 1983) for the path-optimization. With the help of a cost function, which is minimized by the algorithm, it is possible to minimize angular velocities and accelerations, to avoid joint limits, singularities and collisions with obstacles, as well as to find periodic robot trajectories for given periodic effector motions. With a simulation program written in C language this "Constrained Kinematics / Stochastic Optimization" (CKSO) method is tested by application to various tasks that are frequently treated in the literature.

robot grasping through a hole
Fig. 1: Redundant (4 joints) robot grasping through a hole, without colliding with the walls

robot approaching an object at a corner
Fig. 2: Approaching the object at a corner

robot processing an object with corners
Fig. 3: Processing an object with corners; periodic trajectory

You can download the complete paper [Th. Kastenmeier and F. Vesely, "Numerical robot kinematics based on stochastic and molecular simulation methods", Robotica (1996) volume 14 part 3, pp 329-337 (Cambridge University Press, 1996)] in postscript format (813 kB): robpaper.ps

You can also download a PC-executable demo to see the robot arm in action with several different 2D-tasks (188 kB): robidemo.zip

On the 25th of November 1996 we took out a patent for our newly developed robot control-method with the title "Steuerungsverfahren für Roboter, unter Verwendung von geometrischen Zwangsbedingungen und Zufalls-Suchverfahren", patent number AT 401 746 B (we applied for the patent on 13.04.1995). You can download the complete patent specification in MS Word 97 format (163 kB; in German!): patent.doc

If you want to make me happy, leave a nice message in my guestbook!

Home For anyone interested you can see my homepage.

Mail To contact me, please send an email to kast@onlinehome.de


Jump
Back to the top

Keywords: shake simulated annealing Robotica patent robotics robot Roboter arm redundant inverse kinematics Kinematik Steuerung control optimization Optimierung constraint obstacle Hindernis molecular algorithm task stochastic simulation shake simulated annealing Robotica patent robotics robot Roboter arm redundant inverse kinematics Kinematik Steuerung control optimization Optimierung constraint obstacle Hindernis molecular algorithm task stochastic simulation