Shift Modeling Based on Long Short-Term Memory for Manual Transmission System and Shift Process Improvement by Using Bionic Optimization

Shift modeling has increasingly attracted wide attention since it plays a vital role in the intelligent control system for manual transmission (MT) vehicles. In this paper, a modeling scheme based on distal control and real-world vehicle test data (VTD) is proposed to make it possible to imitate the drivers' shift operation. The shift model adapts to be the inverse model of running vehicle, with vehicle speed and engine speed acting as inputs while operation signals as outputs. To tackle the problem of mapping signals from sequence to sequence, the long short-term memory (LSTM) algorithm is adopted, imitating human operations on gas or clutch pedals during shift process. Modeling results reveal that LSTM algorithm obtains satisfying fitting performance. Finally, in terms of fuel economy and driveability, particle swarm optimization (PSO) algorithm is utilized to improve the shifting process. Simulations are conducted to verify the effectiveness of the proposed scheme.