Elsevier

Neurocomputing

Volume 16, Issue 4, 15 September 1997, Pages 333-349
Neurocomputing

Paper
Forward kinematics solution of Stewart platform using neural networks

https://doi.org/10.1016/S0925-2312(97)00048-9Get rights and content

Abstract

The Stewart platform's unique structure presents an interesting problem in its forward kinematics (FK) solution. It involves the solving of a series of simultaneous non-linear equations and, usually, non-unique, multiple sets of solutions are obtained from one set of data. In addition, most effort usually result in having to find the solution of a 16th-order polynomial by means of numerical methods. A simple feed-forward network was trained to recognise the relationship between the input values and the output values of the FK problem and was able to provide the solution around an average error of 1.0 ° and 1.0 mm. By performing a few iterations with an innovative offset adjustment, the performance of the trained network was improved tremendously. Two extra iterations with the offset adjustment reduced the average error of the same trained neural network to 0.017 ° and 0.017 mm.

References (26)

  • T. Arai et al.

    Characteristics and mechanism analysis of parallel link manipulator

  • Z. Geng et al.

    On the dynamic model and kinematic analysis of a class of Stewart platforms

    Robotics Autonomous Systems

    (1992)
  • C.C. Nguyen et al.

    Dynamic analysis of a 6 DOF CKCM robot end-effector for dual-arm telerobot systems

    Robotics Autonomous Systems

    (1989)
  • F. Behi

    Kinematic analysis for a six-degree-of-freedom 3-[PRPS] Parallel mechanism

    IEEE J. Robotics Automation

    (1988)
  • R. Boudreau et al.

    Solving the forward kinematics of parallel manipulators with a genetic algorithm

    J. Robotic Systems

    (1996)
  • K. Cleary et al.

    A prototype parallel manipulator: kinematics, construction, software, workspace results, and singularity analysis

  • W.Q.D. Do et al.

    Inverse dynamic analysis and simulation of a platform type of robot

    J. Robotic Systems

    (1988)
  • P. Fanghella et al.

    Kinematics of robot mechanisms with closed actuating loops

    Int. J. Robotics Res.

    (1990)
  • E.F. Fichter

    A Stewart platform-based manipulator: general theory and practical construction

    Int. J. Robotics Res.

    (1986)
  • Geng Zheng et al.

    Neural network solution for the forward kinematics problem of a Stewart platform

    Robotics Computer-Integrated Manufacturing

    (1992)
  • C. Gosselin et al.

    Singularity analysis of closed-loop kinematic chains

    IEEE Trans. Robotics Automation

    (1990)
  • C. Gosselin

    Stiffness mapping for parallel manipulators

    IEEE J. Robotics Automation

    (1990)
  • H. Inoue et al.

    Parallel manipulator

  • Cited by (84)

    • Validation of a Stewart platform inspection system with an artificial neural network controller

      2022, Precision Engineering
      Citation Excerpt :

      The application of ANNs for Stewart Platforms can be very useful in different ways. As shown in different studies, the modelling of the Stewart Platform dynamics [26] and kinematics [27] can simulate the Stewart Platform accurately. In the modelling process, it is important to design an appropriate ANN structure, i.e., number of neurons in hidden layers, to represent the system successfully [28].

    • Forward and inverse kinematics of a 5-DOF hybrid robot for composite material machining

      2020, Robotics and Computer-Integrated Manufacturing
      Citation Excerpt :

      In general, the direct kinematic equations are a set of nonlinear equations and certain elimination methods such as wu-elimination method [6], Sylvester's dialytic elimination [7] and other special elimination method [8], are utilized to transfer the equations to be a polynomial equation with a single unknown. The most general polynomial equation is an 8th-, 16th- or higher order equation, which in turn has to be solved by some numerical methods such as Taylor series expansion [9], neural networks strategies [10,11], and Newton classical methods [12]. For example, Geng et al. used neural networks to find the forward kinematics solution of Stewart platform [13].

    • Real time direct kinematic problem computation of the 3PRS robot using neural networks

      2018, Neurocomputing
      Citation Excerpt :

      Finally, in the literature very few works analyze the Real-Time performance of DKP solving ANNs [19]. In [14,16], the time performance of the selected MLPs used to solve the DKP of parallel robots is detailed and compared with traditional NR approaches, resulting in a decrease in an order of magnitude on the computational time. However, these measures are carried out with a particular test trajectory, and only consider the final network architecture, and do not analyze the effect of network size in the time performance.

    View all citing articles on Scopus
    View full text