Mobile computing has emerged as an important paradigm to envision the ‘last mile’ of computing services to mobile users. There are many novel distributed computing methods which can be applied for vehicle users to let them cooperatively train ML models for future AI applications, However, traditional centralized training methods are not suitable for vehicle users since they are not connected by a reliable and bandwidth-rich Internet access, which is highly dynamic and often suffers a lot from interruption, interference, etc. In this chapter, we exploit two learning paradigms and analyze their performance based on the IoV connectivity and exploit the vehicles’ mobility. Due to the specific mobility pattern and communication characteristics in IoV, excessive training latency can be caused by the communication bandwidth constraints in vehicular environments, non-negligible volumes of iteration parameters and heterogeneity in computing capacities of distributed workers, etc. We propose novel computing methods to seek the possibility to provision artificial intelligence (AI) to the mobility world with the help of IoV, which has great potential to bring the power of AI to all road users, to support a variety of intelligent applications, e.g., autonomous driving, road safety, ITS, etc.
Intelligent Networking enabled Vehicular Distributed Learning
Internet Access in Vehicular Networks ; Chapter : 5 ; 145-162
2021-09-25
18 pages
Article/Chapter (Book)
Electronic Resource
English
5G-enabled vehicular communications and networking
TIBKAT | 2019
|TIBKAT | 2015
|SLUB | 2014
|SLUB | 2015
|