[3Q4-J-13] AI application: transformation system

[3Q4-J-13-03] Classification for time-sequence data appeared in shift control of automobile automatic transmission

〇Yusuke Morikawa¹, Yasuhiro Ishihara¹, Takanori Ide¹, Eiji Moriyama¹, Taku Akita¹, Yasuo Tabei², Takehito Utsuro³, Hiroshi Nakagawa² (1. Aisin AW Co., Ltd., 2. RIKEN, Center for Advanced Intelligence Project, 3. Fclty. Eng, Inf. & Sys, Univ. of Tsukuba)

Keywords:Support Vector Machine, Time sequence data, Classification, Automatic Transmission, Automobile

The automatic transmission is a transmission that automatically changes gear according to an optimal gear step decided based on driver’s accelerator/brake operation, vehicle speed, etc. Since the quality of a changing gear of automatic transmission, we call shift feeling, has a great influence on riding comfort of the automobile, the parameters of control program are adjusted to improve the shift feeling at the development site. We call this adjustment work calibration. In the process of calibration, there is a process of identifying occurring phenomena in the time-sequence data and selecting the cause and countermeasures. Because this task takes large man-hours, it is very efficient if this task becomes automatic. Therefore, at the first step, in order to automate identification of the occurring phenomena, we built classifier of the time-sequence data using SVM. The experimental result suggests that the proposed identification system is promising directions for actually used identification system.