Driving Intention Identification Based on Long Short-Term Memory and A Case Study in Shifting Strategy Optimization

Abstract

Identification of driving intentions has increasingly attracted wide attention since it can be a valuable reference input of vehicle intelligent control systems. In this study, the long short-term memory (LSTM) is employed to identify the longitudinal intention online with high precision. To this end, the driving intentions when the vehicle runs on a straight and flat road are divided into five categories. The vehicle driving states such as the vehicle speed and acceleration are pre-processed to label the road test data. Subsequently, a LSTM classification model is established to identify the driving intention with inputs of opening degree of the accelerator pedal, vehicle speed and brake pedal force. Identification results reveal that the highest accuracy of the proposed algorithm attains 95.36%, which is around 20% higher than that of the traditional back propagation neural network. Finally, a driving intention-perceptive gear shifting strategy is developed with the help of the built recognition algorithm, and simulation results highlight that the strategy can effectively reduce the number of shifts and achieve better fuel economy.