A Momentum Recurrent Neural Network for Sparse Motion Planning of Redundant Manipulators With Majorization-Minimization
Haoen Huang, Long Jin, Zhigang Zeng
Abstract
In recent decades, despite significant advancements in neural networks for motion planning, the convergence speed is a critical limitation. Additionally, only a few approaches explore the incorporation of sparsity into the motion planning of redundant manipulators. In this article, inspired by Nesterov’s accelerated gradient method, a momentum recurrent neural network (MRNN) model is proposed. For sparse motion planning, multiple MRNNs operate concurrently within a framework of collaborative neurodynamic optimization (CNO). Computer simulations and physical experiments are performed to demonstrate the superiority of MRNN over the existing neural networks in terms of both time efficiency and tracking performance. Specifically, the tracking performance of the proposed CNO-MRNN is the best compared among other competing methods with the maximum position error <inline-formula id="IE1" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><mml:math display="inline"><mml:mrow><mml:mn>2.34</mml:mn><mml:mo>×</mml:mo><mml:msup><mml:mrow><mml:mrow><mml:mn>10</mml:mn></mml:mrow></mml:mrow><mml:mrow><mml:mo>−</mml:mo><mml:mn>3</mml:mn></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> m, mean square error <inline-formula id="IE2" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><mml:math display="inline"><mml:mrow><mml:mn>1.51</mml:mn><mml:mo>×</mml:mo><mml:msup><mml:mrow><mml:mrow><mml:mn>10</mml:mn></mml:mrow></mml:mrow><mml:mrow><mml:mo>−</mml:mo><mml:mn>3</mml:mn></mml:mrow></mml:msup></mml:mrow></mml:math></inline-formula> m, and number of active joints 4.