| 摘要 |
PURPOSE:To reduce the burden on a computer by not only improving the estimation precision but also reducing the number of neurons in an intermadiate layer at the time of estimating a yaw rate by neural network operation. CONSTITUTION:A rotation difference RV between right and left wheels, steering angle Fs, vehicle velocity V, and vehicle body acceleration (g) are detected by detecting means 5, 3, 13, and 16, respectively. After the rotation difference Rv between right and left wheels and the steering angle Fs are normalized, an approximate yaw rate value phip is calculated by a first neural network operation part 8 which takes these normalized values Rvn and Fsn as input variables. After the approximate yaw rate value phip, the vehicle velocity V, and the vehicle body acceleration (g) are normalized, an estimated yaw rate value phi is calculated by a second neural network operation part 10 which takes these normalized values phipn, Vn, and gn as input variables. |
