Analysis and review of multimodal trajectory prediction methods in complex dynamic scenes: evolution from classical models to deep learning

Abstract

This article provides a systematic review of the mainstream algorithms and methods in the field of trajectory prediction for autonomous vehicles, categorizing them into four major approaches: traditional statistical methods, machine learning-based methods, deep learning-based methods, and hybrid models. Through a comprehensive analysis of the principles, strengths, weaknesses, and relevant literature of each approach, this study offers a detailed comparison of their performance characteristics and delves into their respective application scenarios. Furthermore, based on the current state of research, the article explores future directions for trajectory prediction technologies and proposes corresponding research recommendations.