An improved hybrid method for forward kinematics analysis of parallel robots

Abstract

This paper combines a new structure of artificial neural networks (ANNs) with a 3rd-order numerical algorithm and proposes an improved hybrid method for solving forward kinematics problem (FKP) of parallel manipulators. In this method, an approximate solution of the FKP is first generated by the neural network. This solution is next considered as an initial guess for the 3rd-order numerical technique which solves the nonlinear forward kinematics equations and obtains the answer with a desired level of accuracy. To speed up the method, a new structure is proposed for designing the ANN which is called Same Class One Network. In this structure, the outputs of the ANN are classified into classes of similar variables with an individual network designed for each class. The proposed method is then applied to a planar 3-RPR parallel manipulator and a spatial 3-PSP parallel robot. The results show that using this method will lead to a 55% reduction in required iterations and a 20% reduction in the FKP analysis time, while maintaining a high level of solution accuracy.

An Efficient Numerical Method for Forward Kinematics of Parallel Robots

“…Though the simplified Newton iteration saves computing resources in each iteration step, it will weaken the efficiency of the proposed method if the number of iteration is excessive due to a large threshold. Despite this, the required time is still shorter than that in [24] thanks to the initial value and fixed inverse of Jacobian are directly transferred from the previous cycle.…”

“…Therefore, we can obtain the balance between accuracy and efficiency. The research of [24] has made a comparison of ANNs with different structures. Though the structure of ANNs applied in [24] achieves the separate output of position and orientation, it increases the workload during the training phase and is time-consuming because of the extra network.…”

“…The research of [24] has made a comparison of ANNs with different structures. Though the structure of ANNs applied in [24] achieves the separate output of position and orientation, it increases the workload during the training phase and is time-consuming because of the extra network. As a result, the adopted structure of the network in this paper is an all class one network (ACON) with six inputs and six outputs.…”

“…To further verify the improvement of the deviation-driven algorithm in terms of efficiency, NR algorithm, GNRMD algorithm and a 3rd-order numerical algorithm [24] were used as comparisons under the same environment and data. The solution of the previous cycle is used as the initial guess for the current cycle for NR algorithm.…”

Modeling in Two Configurations of a 5R 2-DoF Planar Parallel Mechanism and Solution to the Inverse Kinematic Modeling Using Artificial Neural Network

“…Parallel manipulators have been studied by several researchers over the past two decades, because they have competitive advantages over open-chain robots, for example, greater accuracy, increased load capacity, more rigidity, among others [1]. These features are essential in industrial applications, such as simulators, machine tools, and CNCs, among others [2]. Due to their special characteristics, parallel robots are currently studied by various authors [3,4,5,6,7,8].…”