Evaluation of current vehicle lane-changing prediction methods

• Main Author: Liu, Tian
• Other Authors: Su Rong
• Format: Thesis-Master by Coursework
• Language: English
• Published: Nanyang Technological University 2023
• Subjects: Engineering::Electrical and electronic engineering::Control and instrumentation::Control engineering
• Online Access: https://hdl.handle.net/10356/164982
• Institution: Nanyang Technological University

This dissertation mainly focuses on applying various algorithms to predict vehicle lane changing behavior, an important part of autonomous driving which attempts to achieve intelligent transportation and society. Recently, machine learning methods in the field of artificial intelligence have become mainstream in the direction of human behavior prediction. By training neural network models in supervised or unsupervised learning ways to solve problems in different scenarios, so as to further improve the efficiency of production and life. In order to better compare the performance of existing algorithmic models for lane change behaviour prediction, five network models are trained for the classification task based on the NGSIM I-80 dataset in this dissertation. The most important element for machine learning methods is the data pre-processing, and this dissertation applies a large amount of code to implement feature extraction on the dataset. The datasets are divided to change and keep data to ensure tarining is balanced. Then the training results of LSTM, CNN, LSTM-CNN are compared as a group in this dissertation, and the results of BNN and ANN are as another group. The experimental results are compared to provide a reference for other researchers.