An LSTM-based driving operation suggestion method for riding comfort-oriented critical zone

Driving behavior optimization can not only reduce energy consumption and the probability of traffic accidents but also improve the riding experience of passengers. Unfortunately, the low estimation accuracy resulting from the poor performance of prediction models greatly influences bus service performance. In this paper, a time cycle neural network, the long short-term memory (LSTM) network, is used to evaluate real-time bus riding comfort and provide driving suggestions. To ensure the prediction accuracy, a series of preprocessing procedures, such as data filtering, GPS data processing, parameter calculation and road segmentation, are performed. Three indicators, velocity, longitudinal acceleration, and yaw rate, are selected, while a critical zone-oriented training process is performed. Simulation results show that the proposed method has rapid convergence and acceptable prediction accuracy while providing driving suggestions is reasonable.