On-road feature detection and fountain-coded data dissemination in vehicular ad-hoc networks

Abstract

Within the smart city framework, information dissemination in vehicular ad-hoc networks (VANET) is attracting considerable interest in both the research community and industry. Efficient data dissemination has long been a problem in ad-hoc networks. In VANET, the problem is even more challenging given the high mobility of vehicles, high density of buildings, and intermittent network connectivity. Realistic modelling of the mobility patterns of vehicles (instead of random models like Random Waypoint or Manhattan Models in previous works) is important for accurate performance evaluation. In this paper, the transmission of on-road feature detection data (images or videos) with fountain code in VANET is studied. Specifically, we propose a robust license plate detection module and applied fountain coding in the application layer to largely reduce the average transmission delay of multi-media data. The proposed system is rigorously evaluated under a semi-realistic simulation of an inter-bus communication network in the Mong Kok urban district in Hong Kong with the consideration of real-world traffic parameters, such as different traffic density at different hours of a day and building obstacles. Specifically, we find that fountain coded data dissemination shows better performance boost in more realistic signal propagation model in urban areas. In practice, the proposed system can be applied to bus lane occupancy control. For example, when a vehicle illegally occupies the bus lane, buses nearby can help recognize the vehicle and transmit the detected results through the VANET to patrol cars for the enforcement.