An improved neural network based fuzzy self-adaptive KALMAN filter and its application in cone picking robot

• Published in: 2009 International Conference on Machine Learning and Cybernetics Vol. 1; pp. 573 - 577
• Main Authors: Xiu-Rong Guo, Feng-Hu Wang, Dan-Feng Du, Xiu-Li Guo
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2009
• Subjects: Automatic control; Cone picking robot; Control systems; Cybernetics; Forestry; Fuzzy neural networks; Fuzzy self-adaptive KALMAN filter; Hydraulic drive; Machine learning; Manipulators; Neural networks; RBFNN (radial basis function neural network) controller; Robotics and automation; Service robots
• ISBN: 9781424437023; 1424437024
• ISSN: 2160-133X
• DOI: 10.1109/ICMLC.2009.5212508

Aimed to improve the working efficiency of cone picking robot and release workers from heavy manual labor, a novel RBF neural network based fuzzy self-adaptive Kalman filter is presented in the paper. The position and object input voltage are taken as the inputs of the RBF neural network model. Consider that the traditional BP algorithm has shortcomings of converging slowly and easily trapping a local minimum value, a combination learning algorithm using fuzzy self-adaptive Kalman filter is adopted to train the neural network. The sample data obtained from the 3D laser scanner and sensors located on the cone picking robot. Experimental results show that it will enable the training process with an overall accuracy and rapid convergence speed. The application of the technology in cone picking robot automatic control system proves it is an effective method and has certain project value.