Automobile Technology

Research on Identifica­tion of DCT Vehicle Driver’s Starting Intention Based on GM-HMM Liu Haijiang, Su ………………………………………………

- Key words: DCT starting process, Starting intention, GM-HMM

Liu Haijiang, Su Bowei Tongji University, Shanghai 201804) ( Abstract The current DCT control system has low accuracy in the identifica­tion of driver's starting intention. To【 】solve this problem, a starting intention identifica­tion method based on Gaussian Mixed- Hidden Markov model (GM- HMM) is proposed: according to the longitudin­al accelerati­on distributi­on characteri­stics of the DCT vehicle, the starting process is divided into 8 periods. The smooth starting, general starting and emergency starting are defined based on the K- means clustering algorithm. On this basis, the 3 types of GM-HMM are trained in each period, and the current starting intention of the driver is determined by comparing the log- likelihood ratios of the accelerato­r pedal opening time sequence in 0.3 s in different models. The research result shows that the average recall ratio of the identifica­tion model reaches 88.7%, and it only takes 7 ms, which has higher recognitio­n accuracy and better real-time performanc­e.

Newspapers in Chinese (Simplified)

Newspapers from China