۱- Introduction

۲- Related work

۳- Preliminaries

۴- Filter-based efficient request answering

۵- Update a single filter

۶- Filter cube construction and update

۷- Experimental study

۸- Conclusions

References

Abstract

Vehicles in urban city are equipped with more and more sensing units, and sensed data are continuously generated in large amount. These sensed data could be filtered and preprocessed before being shared or uploaded to the road side units and the cloud for efficiency. In this paper we propose a filter-based framework called FERA (Filter-based Efficient Request Answering), which combines the concept of fog computing and vehicular sensing, and adopts the pull/push strategies to adaptively and efficiently gather the requested data in vehicular ad hoc networks. Filters are defined based on the ratio of cost between the push and the pull methods to control the passage or blockage of the data readings. Moreover, filter cubes are defined to manage large number of filters, where efficient algorithms are developed to construct, update and store the filter cubes so that the matched data readings are pushed upward and unmatched data readings are blocked effectively. Extended simulated experiments demonstrate the proposed scheme has a much higher success ratio of request answering than existing schemes, e.g. REED (Abadiet al., 2005) and GeoVanet (Delotet al., 2011). Up to 94 percent of the requests could be successfully processed, while at the same time maintaining a relatively low query cost.

Introduction

Vehicular nodes are equipped with more and more sensing units, and large amount of sensing data such as GPS locations, speeds, video clips, and so on are generated [1,2]. These data are shared or uploaded as inputs for applications that aim at more intelligent transportations, emergency responses, and reduced pollutions and fuel consumptions. So cooperative urban sensing [3,4] is at the heart of the intelligent and green city traffic management. The key components of the platform will be a combination of pervasive vehicular ad hoc network and a central control and analyzing system. This has led to the emergence of a new kind of system, i.e. the Vehicular Ad-hoc Sensing System [5,6], where vehicles travel along roads and exchange information with encountered vehicles or nodes through V2V(vehicle to vehicle) or V2I (vehicle to infrastructure) communications. Data can be disseminated and reach a far distance by using moving vehicles as intermediates, following multi-hop routing protocols. Recently, IEEE 802 committee defined wireless communication standard IEEE 802.11p [7] that serves specifically for V2I communication. The Federal Communications Commission has allocated 75 MHz of bandwidth, which operates on 5.9 GHz channel for short range communications. One key and challenging issue in VANET is the vehicular data gathering [2,8–۱۰]. First, vehicular nodes are limited to road topology while moving, and under various road conditions and high moving speeds the network usually suffers rapid topology and density changes. The communications are usually fragmented and intermittent-connected. Second, the vehicular sensed data is in large amount and characterized as continuous generation. The sensed data should be filtered and preprocessed before being shared or uploaded. Data filtering technologies tailored to the VANET environment are highly needed. Generally speaking, there are two strategies to gather data: the push-based and the pullbased models, which are similar to those strategies used in the field of distributed and mobile databases. In a push-based model, each vehicle senses the data and proactively to upload data to a central server through V2V or V2I communications [11,12]. So when a node receives information from its neighbors, it has to decide whether that information is relevant or not.