SEMI-INFINITE PROGRAMMING TO SOLVE ARMED ROBOT TRAJECTORY PROBLEM USING RECURRENT NEURAL NETWORK

Alaeddin Malek, Leila Jafarian-Khaled Abad, and Samaneh Khodayari-Samghabadi

Keywords

BSpline, semiinfinite programming, discretization method, recurrent neural network.

Abstract

In this paper, an armed robot trajectory planning problem in the robot joint space is considered. A model for finding the best path with minimum time is proposed. The relation between robot joint space path and time using B-spline curve is constructed. The modelling is lead to semi-infinite programming (SIP) problems. Discretization method over SIP is used to solve the problem successfully. The optimization problem by replacing the continuous interval of parameter joint space path with subset of discrete point of the interval is proposed. The corresponding optimization problem with specific recurrent neural network (RNN) model is solved. The current scheme is independent of starting point and finds efficiently both primal and dual solution simultaneously. The optimal values of this problem are the control points that show the position of the armed robot trajectory to take the lowest possible time. Numerical simulation illustrates efficiency of this model.

Important Links:



Go Back