Hybrid Control of Robotic Manipulator by Neural Network Model : 4th Report, Sensing and Hybrid Control with Collision Phenomena

• Published in: Transactions of the Japan Society of Mechanical Engineers Series C Vol. 57; no. 539; pp. 2305 - 2312
• Main Authors: SHIBATA, Takanori, FUKUDA, Toshio, ARAI, Fumito, WADA, Hiroshi, TOKITA, Masatoshi, MITSUOKA, Toyokazu, SHOJI, Yasumasa
• Format: Journal Article
• Language: Japanese
• Published: The Japan Society of Mechanical Engineers 25.07.1991
• Subjects: Collision; Dynamics; Force Control; hybrid Control; Manipulator Control; Neural Network; Robotics; Sensing
• ISSN: 0387-5024; 1884-8354
• DOI: 10.1299/kikaic.57.2305

This paper deals with neural network applications for robotic motion control where the controller is applicable to the position and force control of robotic manipulators undergoing collisions. Collision phenomena have strong nonlinearity and are high-speed phenomena, so it is difficult to sense collision phenomena and to control robotic manipulators undergoing collisions. The neural servo controller is based on the neural network which has integrated time delay elements, so that the neural network can learn the nonlinear dynamical system. Therefore, the neural servo controller is effective against collision phenomena. We adopted a Hertz-type model with an energy loss parameter to express the impact force between the manipulator and unknown objects. Simulations are carried out for the case of hybrid control of a one-dimensional robotic manipulator which handles unknown objects using proximity sensors. The results show that the neural servo controller can sense and control collision phenomena. Moreover, they show the relation between the intelligence of the neural network and information from sensors.