A moving time node method for solving robot optimal control problem
Enhui Jiang, Weifeng Xu, Aipeng Jiang, Hanhan Gao, Jian Wang, Yangiian Xiao · 2019
Many motion control problems of robots are typical Bang-Bang optimal control problems with multi-control variables. In order to solve the contradiction between discrete approximation accuracy and computational time, and the inaccurate location of time-switching points caused by fixed time grids, a method of moving time nodes is proposed for more reasonably dividing the discrete-time grids of optimal control problems. In this method, each component of the control variable is taken as the parameter to be optimized, and the time grid distribution of each control variable is obtained. Unlike the Time-Scaling method, this method not only can effectively move the time grid nodes and obtain good time grid distribution, but also has stronger approximation ability and more flexibility. The effectiveness of the proposed method is verified by a robot optimal control example, and compared with the traditional control vector parameterization method and Time-Scaling method. The results show that the proposed method has low computational cost, accurate time-switching point location and high accuracy.