Structure designing of BP neural network in the application of reference velocity estimation

Author

Guirong Zhuo ; Bingxue Wang

Author_Institution

Clean Energy Automotive Eng. Center, Tongji Univ., Shanghai, China

fYear

2014

fDate

3-6 Aug. 2014

Firstpage

1481

Lastpage

1485

Abstract

BP neural network (BPNN) is used to estimate vehicle velocity when car brakes and ABS functions. Based on Fuzzy C-Means (FCM) clustering algorithm, a new empirical formula of hidden-layer nodes is proposed. Adding delays to the input-layer of BPNN for expanding the input sample space can improve estimated accuracy greatly. The appropriate distributed delays selected can reduce the redundancy of the network structure, and improve the mapping relationships of the inputs and outputs. Velocity estimation is simulated on the condition of high adhesion-coefficient road, and the results show that the absolute error is no more than 1 km/h and the relative error is no more than 0.4%.

Keywords

automobiles; backpropagation; brakes; braking; delays; distributed control; fuzzy control; neurocontrollers; pattern clustering; velocity control; ABS functions; BP neural network; BPNN; FCM clustering algorithm; absolute error; anti-lock braking system; car brakes; distributed delays; fuzzy c-means clustering algorithm; hidden-layer nodes; high adhesion-coefficient road; mapping relationships; network structure redundancy; reference velocity estimation; structure designing; vehicle velocity estimation; Accuracy; Delays; Estimation; Neural networks; Training; Vehicles; Wheels; BP Neural Network; Fuzzy C-means Clustering; Input Delays; Redundancy of Network Structure; Vehicle Velocity Estimation;

fLanguage

English

Publisher

ieee

Conference_Titel

Mechatronics and Automation (ICMA), 2014 IEEE International Conference on

Conference_Location

Tianjin

Print_ISBN

978-1-4799-3978-7

Type

conf

DOI

10.1109/ICMA.2014.6885918

Filename

6885918