Skip to main content

Multimedia Communication and Processing over Vehicular Cloud Networks for AutonomousDriving

Periodic Reporting for period 1 - MACROSS (Multimedia Communication and Processing over Vehicular Cloud Networks for AutonomousDriving)

Reporting period: 2018-03-01 to 2020-02-29

Since the invention, vehicles have significantly improved human’s life quality, but also brought serious traffic problems. Integrating information and communication technologies (ICT) into the transportation infrastructure to establish intelligent transportation system (ITS) has been deemed as an inevitable trend in the evolution of modern transportation and traffic management. Several companies and research labs have successfully demonstrated prototypes of automated cars. However, a great number of theoretical and technical challenges still need to be addressed in order to reap the full potential of ITS. For instance, the decision-making process of most current driver assistance applications and autonomous driving functions completely rely on the processing of on-board sensor data. Due to the physically constrained sensing range and accuracy, as well as limited computation and storage capacity, a single vehicle’s capability of understanding the complex driving environment can be unsatisfactory. An effective solution to this issue is to allow the sensing/computing/storage resources of different elements in the ITS to be shared, so that the environment perception and decision-making ability of each individual can be greatly enhanced. This idea stems from the cloud computing technology and constitutes a recently emerged concept, vehicular cloud network (VCN). Supported by the mobile edge computing (MEC) framework, VCN enables ITS services and applications that individual vehicles cannot produce alone.

High-quality wireless communication that guarantees reliable real-time sensing data (normally in the form of multimedia) to be shared among vehicles and road-side infrastructure is the key to realize VCN. The vehicle-to-everything (V2X) communication technology has attracted tremendous research attentions. This motivates another technical path towards fully autonomous driving by improving the environment perception capability of smart cars with reduced cost through V2X communications. However, supporting future autonomous driving applications with shared sensing data is challenging. On the one hand, transmission protocols are designed mainly targeting on small-size bursty emergency-triggered messages or periodic vehicle-status messages. On the other hand, investigations exploiting the full potential of sharing content-rich and resource-hungry sensing data normally focus on effective data processing based on the assumptions that the wireless data delivery process is either sufficiently good or completely unpredictable. These designing perspectives would lead to unsatisfactory outcomes. To this end, the project aims to develop the effective methods that leverage innovative multimedia communication and multimedia processing technologies over VCNs to support future autonomous driving. It targets to tackle two challenges: 1) how to transmit resource-hungry multimedia sensing data in unreliable vehicular communication environments and maximize the data sharing quality; and 2) how to process the sensing data when only a certain level of knowledge regarding the network communication quality is available.
We have developed several novel transmission protocols based on cooperative communication and cross-layer transmission design techniques. They have been shown to be able to improve the reliability and efficiency of multimedia sensing data transmission among vehicles over complex fading channels, compared with existing solutions. We have also designed new methods based on mathematical tools and machine learning techniques to determine how to process complicated on-board application tasks through shared computation resources in uncertain dynamic environments, when the wireless conditions can only be partially available. We have implemented two testbed experimental systems and conducted field-experiments to demonstrate the functionality of VCNs in supporting extending the sensing range of individual vehicles and also supporting autonomous driving applications. The methods and results achieved in the project can serve as theoretic tools and potential technical solutions for supporting future autonomous driving functions over VCNs. Extending the research works of the project from different perspectives and further exploiting the findings are being discussed among the fellow, the hosting group, and their industrial partners.
The number of worldwide car sold has steadily increased for decades. According to the Beijing Traffic Management Bureau, Beijing had 6.3 million motor vehicles by March 2020, which was an increase of 4.8 million since 2000. Such a huge number of vehicles running on the roads have significantly improved human’s life quality, but also brought a series of traffic accident and jam problems, which cause millions of life casualties, unquantifiable wasted time/capitals/energy, and serious global pollution issues. For instance, the Global Status Reports on Road Safety launched by the World Health Organization (WHO) reported that road traffic crashes had caused over 1.25 million deaths annually and become the leading killer of young people. Although many governments and automobile manufacturers are continuously working on constructing new roads and bridges, optimizing traffic rules, and developing better automobile life-protection and navigation technologies, it is widely accepted that conventional traffic management and automobile engineering approaches cannot completely solve these issues. Integrating advanced sensing, communication, and computing technologies to enable ITS is widely believed to be the solution. As a key element in ITS, smart autonomous driving cars will play a key role for reducing traffic accident, jam rates, environment pollution, and the burden of drivers. However, although a number of recent high-profile successful experimental results may give the impression that the day fully-functional autonomous cars running on the roads would come soon, in fact numerous scientific and technical challenges still await to be solved, since actions and decisions in transportation systems after all are related to human life and hence must be taken with caution. In this project, we have developed several novel methods to support efficient and reliable sensing data transmission and processing over VCNs. The results achieved in the project can serve as theoretic tools and potential technical solutions for supporting the development of future autonomous driving functions to address traffic problems, and hence bring benefits for the society.
A VCN