LEARNING TYPE VECTOR OPERATION UNIT

摘要

PROBLEM TO BE SOLVED: To improve the adaptivity to environmental change in decentralized vector operation processing using a sine function type wave, and the robustness of the operation as to an operation technique for vectors which is used for a positioning system for navigation of a moving body, etc. SOLUTION: An azimuth detection unit 101 calculates an output value Xj corresponding to an inputted azimuthϕ. A cos conversion part 102 calculates a sine function type output waveform value Yi corresponding to the output value Xj . A speed signal part 104 imposes amplitude modulation on the output waveform value 103 with a moving speed (r). A short-period memory 103 performs vector addition of the output waveform value Yi which is inputted sequentially after the amplitude modulation. A 1st learning device 103 updates a transformation matrix Sij of the cos transformation part 102 by learning. A 2nd learning device 104 updates a combination matrix Tij of the short-period memory 103 by learning.