Towards a Data-Driven Digital Twin AI-Based Architecture for Self-Driving Vehicles

Abstract

Recent advancements on digital technologies, particularly artificial intelligence, have been resulted into remarkable transformations in automobile industry. One of these technologies is artificial intelligence (AI). AI plays a key role in the development of autonomous vehicles. In this paper, the role of AI in autonomous vehicle (AV) platform layers is studied. The focus of this paper is on the indexed papers in Scopus database. The most relevant keywords are selected and searched. 628 articles, between 2014 and 2024 were selected for analysing and reviewing. Articles were analysed based on source type, topics, and AI algorithms. Text mining and content analysis of articles revealed that 233 journals published 628 articles, and the most top 185 are selected to assess. The topics of paper are classified into perception, localization and mapping, planning, decision making, control, communication, security, data management, and general topics. Each of these areas consisted of many roles, or tasks and use AI to realize their tasks. Convolutional neural network in the perception, control, and localization and mapping layers have been more used. Deep reinforcement learning had the most application in planning and decision-making areas. The main result of this paper is recognition of AVs platform layers classification, designing a data-driven digital twin AI-based model of autonomous vehicles architecture, containing physical world, virtual world, and communication space, and mapping of applied AI algorithms each layer, which aid researchers to choose the suitable methods in the field of autonomous vehicles. This study provided a comprehensive map of research projects related to from 1985 to 2022. Finally, some research directions are suggested.