AUTHORS: Yutaka Maeda, Naoyuki Ishibashi
Download as PDF
ABSTRACT: Robots are widely used in many fields. It is important to provide many different methodologies for robot control. This paper proposes a real time scheme for robots control and learning using recurrent neural network. We handle a problem to control a position and a trajectory of tip of a Selective Compliance Assembly Robot Arm(SCARA) robot. We adopt the simultaneous perturbation optimization method as a learning rule of the recurrent neural networks(RNNs). Then the RNNs have to learn an inverse dynamics of the SCARA robot. Position and trajectory control of a SCARA robot using RNN are considered. We could confirm that the RNNs can learn the inverse dynamics and work as a neuro-controller. We describe details of the control scheme. Some experimental results for these control using an actual SCARA robot are shown
KEYWORDS: Robot control, Learning, Recurrent neural networks, Simultaneous perturbation, SCARA, Inverse dynamics, Real time control
REFERENCES:
[1] S. Yildirim, Robot trajectory control using neural networks, Electronics Letters, 38, 2002, pp. 1111-1113.
[2] J. Reyes-Reyes, C. M. Astorga-Zaragoza, M. Adam-Medina, and G. V. Guerrero-Ramirez, Bounded neuro-control position regulation for a geared DC motor, Engineering Applications of Artificial Intelligence, 23, 2010, pp. 1398–1407.
[3] A. M. Shahri, B. J. Evans, and F. Naghdy, Neuro-fuzzy adaptive torque control of a SCARA robot, 1996 Australian New Zealand Conference on Intelligent Information Systems, 1996, pp. 241-244. 㻜 㻜㻚㻝 㻜㻚㻞 㻜㻚㻟 㻜㻚㻠 㻜㻚㻡 㻜㻚㻢 㻜㻚㻣 㻜 㻝㻜㻜 㻞㻜㻜 㻟㻜㻜 㻠㻜㻜 㻡㻜㻜 㻢㻜㻜 㻣㻜㻜 㻤㻜㻜 㻥㻜㻜 㻝㻜㻜㻜 Iteration Evaluation Figure 24: Change of evaluation function for two trajectories using practical system
[4] S. Omatu, M. Khalid and R. Yusof, NeuroControl and Its Applications, Springer 1996.
[5] N. Ishibashi and Y. Maeda, Learning of InverseDynamics for SCARA Robot, SICE Annual Conference 2011, 2011, pp. 1300–1303.
[6] J. C. Spall, Multivariable stochastic approximation using a simultaneous perturbation gradient approximation, IEEE Trans. Autom. Control, 37, 1992, pp. 332–341.
[7] J. C. Spall, Introduction to Stochastic Search and Optimization, John Wiley & Sons, Inc., 2003.
[8] Y. Maeda, Y. Kanata, A learning rule of neural networks for neuro-controller, Proceedings of the 1995 World Congress of Neural Networks, 2, 1995, pp.402–405.
[9] Y. Maeda, H. Hirano and Y. Kanata, A learning rule of neural networks via simultaneous perturbation and its hardware implementation, Neural Networks, 8, 1995, pp. 251–259.
[10] Y. Maeda, R.J.P.de Figueiredo, Learning rules for neuro-controller via simultaneous perturbation, IEEE Transaction on Neural Networks, 8, 1997, pp.1119–1130.
[11] Y. Maeda, T. Tada, FPGA implementation of a pulse density neural network with learning ability using simultaneous perturbation, IEEE Trans. on Neural Networks, 14, 2003, pp.688-695.
[12] Y. Maeda and M. Wakamura, Simultaneous Perturbation Learning Rule for Recurrent Neural Networks and Its FPGA Implementation, IEEE Trans. on Neural Networks, 16, 2005, pp.1664- 1672.